mldata-utils-0.4.1/0000700000175000017500000000000011477311160013131 5ustar sonnesonnemldata-utils-0.4.1/ml2h5/0000700000175000017500000000000011477311160014060 5ustar sonnesonnemldata-utils-0.4.1/ml2h5/indexsplit.py0000600000175000017500000000700411475246572016634 0ustar sonnesonneimport numpy import re import random def check_split_str(split_str): if re.match("((^|,)(\d+:\d+|\d+))+$",split_str)==None: return False tk=[] tk_sort=[] tk_pr=[] tk_po=[] split_str=str(split_str) tk_split=re.findall("(^|,)(\d+:\d+)|(\d+)",split_str) for i in range(len(tk_split)): if tk_split[i][2]!='': tk.append(tk_split[i][2]) tk_pr.append(int(tk[-1])) tk_po.append(int(tk[-1])+1) if tk_split[i][1]!='': tk.append(tk_split[i][1]) tk_pr.append(int(tk[-1].split(":")[0])) tk_po.append(int(tk[-1].split(":")[1])) tk=numpy.array(tk) tk_pr=numpy.array(tk_pr) tk_po=numpy.array(tk_po) sindex=tk_pr.argsort() tk=tk[sindex] tk_pr=tk_pr[sindex] tk_po=tk_po[sindex] for i in range(len(tk)): if tk_pr[i]>=tk_po[i]: return False return True def expand_split_str(split_str): tk=[] tk_sort=[] tk_pr=[] tk_po=[] split_str=str(split_str) tk_split=re.findall("(^|,)(\d+:\d+)|(\d+)",split_str) for i in range(len(tk_split)): if tk_split[i][2]!='': tk.append(tk_split[i][2]) tk_pr.append(int(tk[-1])) tk_po.append(int(tk[-1])+1) if tk_split[i][1]!='': tk.append(tk_split[i][1]) tk_pr.append(int(tk[-1].split(":")[0])) tk_po.append(int(tk[-1].split(":")[1])) tk=numpy.array(tk) split_pr=numpy.array(tk_pr) split_po=numpy.array(tk_po) out_set = set() for l in range(len(split_pr)): for i in range(split_po[l]-split_pr[l]): out_set.add(split_pr[l]+i) out = [i for i in out_set] out.sort() out_str=[str(i) for i in out] return out_str def reduce_split_str(split_str): tk=[] tk_sort=[] tk_pr=[] tk_po=[] if type(split_str) in [list,numpy.ndarray,numpy.matrix]: split_str=', '.join([str(i) for i in split_str]) else: split_str=str(split_str) tk_split=re.findall("(^|,)(\d+:\d+)|(\d+)",split_str) for i in range(len(tk_split)): if tk_split[i][2]!='': tk.append(tk_split[i][2]) tk_pr.append(int(tk[-1])) tk_po.append(int(tk[-1])+1) if tk_split[i][1]!='': tk.append(tk_split[i][1]) tk_pr.append(int(tk[-1].split(":")[0])) tk_po.append(int(tk[-1].split(":")[1])) tk=numpy.array(tk) tk_pr=numpy.array(tk_pr) tk_po=numpy.array(tk_po) sindex=tk_pr.argsort() tk=tk[sindex] tk_pr=tk_pr[sindex] tk_po=tk_po[sindex] if len(tk) == 1: return tk tk_reduce=[] akt=0 red_ind=0 aktstr=str(tk_pr[0]) for i in range(1,len(tk)): if (tk_pr[i]==tk_po[i-1]): akt+=1 else: if akt>0: aktstr+=":"+str(tk_po[i-1]) tk_reduce.append(aktstr) aktstr=str(tk_pr[i]) akt=0 else: tk_reduce.append(tk[i-1]) aktstr=str(tk_pr[i]) if i==len(tk)-1: if akt>0: aktstr+=":"+str(tk_po[i]) tk_reduce.append(aktstr) aktstr=tk_pr[i] else: tk_reduce.append(tk[i]) return tk_reduce def main(): print check_split_str("1,2,3,4,5:7,27,28:32,32:30") x = reduce_split_str("1") print x y= expand_split_str("1,2,3:7,27:33") print y if __name__=="__main__": main() mldata-utils-0.4.1/ml2h5/__init__.py0000600000175000017500000000034411473473174016206 0ustar sonnesonne""" Utility functions around HDF5, mostly for mldata.org. """ __all__ = ['task', 'data', 'fileformat', 'converter'] VERSION_MLDATA = '0' NUM_EXTRACT = 10 # maximum value length LEN_EXTRACT = 6 PROGRAM_NAME='ml2h5' VERSION='0' mldata-utils-0.4.1/ml2h5/fileformat.py0000600000175000017500000001312711475246572016604 0ustar sonnesonneimport h5py import os.path from ml2h5.converter.basehandler import ALLOWED_SEPERATORS from ml2h5.converter import AUTODETECTION_MAXBUFLEN def _try_suffix(fname): """Get format of given file by suffix. @param fname: name of file to determine format for @type fname: string """ suffix = fname.split('.')[-1] # just assume libsvm if no proper suffix if suffix.find('/') != -1: return '', False if suffix in ('svm', 'libsvm', 'light', 'svmlight'): return 'libsvm', True elif suffix in ('arff'): return 'arff', True elif suffix in ('h5', 'hdf5'): return 'h5', True elif suffix in ('csv', 'tsv'): return 'csv', True #elif suffix in ('uci', 'data'): # return 'uci', True elif suffix in ('bz2', 'gz', 'zip'): try: presuffix = fname.split('.')[-2] if presuffix == 'tar': return presuffix + '.' + suffix, True except IndexError: pass return suffix, True elif suffix in ('mat'): return 'matlab', True elif suffix in ('octave'): return 'octave', True elif suffix in ('xml'): return 'xml', True else: # unknown return suffix, False def _try_arff(fname): """Try if given file is in arff format @param fname: name of file to determine format for @type fname: string """ try: import arff arff.ArffFile.load(fname) return True except: return False def _try_csv(fname): """Try if given file is in csv format @param fname: name of file to determine format for @type fname: string """ if infer_seperator(fname) == ',': return True else: return False def _try_libsvm(fname): """Try if given file is in libsvm format @param fname: name of file to determine format for @type fname: string """ try: fp = open(fname, 'r') except: return False line = fp.readline(AUTODETECTION_MAXBUFLEN) attributes = line.split() if len(attributes) > 1: # 0th might be label, so look at 1st if len(attributes[1].split(':')) == 2: fp.close() return True fp.close() return False def _try_h5(fname): """Try if given file is in hdf5 format @param fname: name of file to determine format for @type fname: string """ try: fp = h5py.File(fname, 'r') except: return False fp.close() return True def _try_matlab(fname): try: return file(fname).read(6) == 'MATLAB' except: return False def _try_octave(fname): try: return file(fname).read(13)==('# Created by ') except: return False def infer_seperator(fname): """Infer seperator for variables in given file. @param fname: filename to retrieve data from @type fname: string @return: inferred seperator @rtype: string """ try: fp = open(fname, 'r') except: return None seperator = None minimum = 1 for i in xrange(100): # try the first 100 lines line = fp.readline(AUTODETECTION_MAXBUFLEN) if not line: break for s in ALLOWED_SEPERATORS: l = len(line.split(s)) if l > minimum: minimum = l seperator = s if seperator: break # stop processing if lines are too long anyways if not line.endswith('\n'): break fp.close() return seperator def get(fname, skip_suffix=False): """Get format of given file. By suffix it detects: libsvm, arff, csv, h5, tar.gz, tar.bz2, zip, matlab, octave. By deeper inspection it detects: h5, matlab, octave, arff, csv, libsvm @param fname: name of file to determine format for @type fname: string @param skip_suffix: if detection by suffix (first priority) shall be skipped @type skip_suffix: boolean """ if not skip_suffix: format, found = _try_suffix(fname) if found: return format if _try_h5(fname): return 'h5' elif _try_matlab(fname): return 'matlab' elif _try_octave(fname): return 'octave' elif _try_libsvm(fname): return 'libsvm' elif _try_csv(fname): return 'csv' elif _try_arff(fname): return 'arff' return 'unknown' def get_filename(orig): """Convert a given filename to something that indicates HDF5. @param orig: original filename @type orig: string @return: HDF5-ified filename @rtype: string """ return os.path.splitext(orig)[0] + '.h5' def can_convert_h5_to(dst_type, h5_filename=None): """Whether conversion from this particular h5 file to dst type is supported @param h5_filename: h5 filename or None @type h5_filename: string @param dst_type: name of type @type dst_type: string @return: True if possible """ if dst_type in ('matlab', 'octave'): return True if h5_filename and h5_filename.endswith('.h5'): try: h5 = h5py.File(h5_filename, 'r') if dst_type=='libsvm': # libsvm requires data/label ordering=set(('label','data')) if ordering.issubset(set(h5['data'].keys())): return True # TODO check if this is sparse data / ndarray data elif dst_type in ('csv', 'arff'): # csv/arff support everything except sparse data for k in h5['data'].keys(): if k.endswith('_indptr') or k.endswith('_indices'): return False return True except: pass return False mldata-utils-0.4.1/ml2h5/data.py0000600000175000017500000003212211473473174015357 0ustar sonnesonne""" Handle Data objects and files. This module heavily relies on the functionality required for http://mldata.org """ import h5py, numpy, os, copy, tarfile, zipfile, bz2, gzip from scipy.sparse import csc_matrix import fileformat, converter from . import NUM_EXTRACT, LEN_EXTRACT def get_num_instattr(fname): """Retrieve number of instances and number of attributes from given HDF5 file. @param fname: filename to retrieve data from @type fname: string @return: number of instances and number of attributes @rtype: tuple containing 2 integers """ try: h5 = h5py.File(fname, 'r') num_inst = 0 num_attr = 0 for name in h5['/data_descr/ordering']: vname = '/data/' + name sp_indptr = vname+'_indptr' sp_indices = vname+'_indices' if sp_indptr in h5['/data'] and sp_indices in h5['/data']: A = csc_matrix( (h5[vname], h5[sp_indices], h5[sp_indptr]) ) num_inst += A.shape[1] num_attr += A.shape[0] else: if len(h5[vname].shape) == 2: num_attr += h5[vname].shape[0] num_inst = h5[vname].shape[1] else: num_inst = h5[vname].shape[0] num_attr += 1 h5.close() except: num_inst = -1 num_attr = -1 return (num_inst, num_attr) def _get_extract_data(h5): """Get extract data from given HDF5 object. @param h5: HDF5 file to get extract from @type h5: HDF5 file @return: data extract @rtype: list """ overlength=False overwidth=False cur_line=0 extract = [] try: for dset in list(h5['/data_descr/ordering']): if cur_line==NUM_EXTRACT: overwidth=True break dset = '/data/' + dset dset_indptr = dset+'_indptr' dset_indices = dset+'_indices' if dset_indptr in h5['/data'] and dset_indices in h5['/data']: # taking all data takes to long for quick viewing, but having just # this extract may result in less columns displayed than indicated # by attributes_names # reduce columns indices = h5[dset_indices][:h5[dset_indptr][NUM_EXTRACT+1]] data = (h5[dset][:h5[dset_indptr][NUM_EXTRACT+1]]) indptr = h5[dset_indptr][:NUM_EXTRACT+1] csc_shape = [min(NUM_EXTRACT,max(indices)),min(NUM_EXTRACT,len(indptr))] if len(indptr) NUM_EXTRACT): overwidth=True # reduce rows for i in xrange(len(indptr)-1): for j in range(indptr[i],indptr[i+1]): if indices[j]>=NUM_EXTRACT: indptr[i+1:]-=numpy.ones(len(indptr[i+1:])) data=data[indices < NUM_EXTRACT] indices=indices[indices < NUM_EXTRACT] # empty array exemption if data.shape==(0,) or indices.shape==(0,): data=numpy.array([0]) indices=data A=csc_matrix((data, indices, indptr),csc_shape).todense() if A.shape[0]+cur_line > NUM_EXTRACT: extract.extend(A[:-cur_line].tolist()) else: extract.extend(A.tolist()) cur_line+=A.shape[0] else: if type(h5[dset][0]) == numpy.ndarray: last = len(h5[dset][0]) if last > NUM_EXTRACT: last = NUM_EXTRACT overlength=True app_lines=xrange(len(h5[dset])) if len(app_lines) + cur_line > NUM_EXTRACT: app_lines = xrange(NUM_EXTRACT - cur_line ) overwidth=True for i in app_lines: extract.append(h5[dset][i][:last]) cur_line+=len(app_lines) else: cur_line+=1 last = len(h5[dset]) if last > NUM_EXTRACT: last = NUM_EXTRACT overlength=True extract.append(h5[dset][:last]) if overwidth: extract.append(['...' for i in extract[0]]) extract = numpy.matrix(extract).T # convert from numpy array to list, if necessary t = type(extract[0]) if t == numpy.ndarray or t == numpy.matrix: extract = [y for x in extract for y in x.tolist()] if overlength: extract.append(['...' for i in extract[0]]) # cut all values if necessary for l in range(len(extract)): for c in range(len(extract[l])): extract[l][c]=str(extract[l][c]) if len(extract[l][c])>LEN_EXTRACT: if extract[l][c][0]=='-': extract[l][c]=extract[l][c][:LEN_EXTRACT-1]+'...' else: extract[l][c]=extract[l][c][:LEN_EXTRACT-2]+'...' except KeyError: pass except ValueError: pass except IndexError: pass return extract def get_extract(fname): """Get an extract of an HDF5 Data file. @param fname: filename to get extract from @type fname: string @return: extract of an HDF5 file @rtype: dict with HDF5 attribute/dataset names as keys and their data as values """ format = fileformat.get(fname) if fname.endswith('.h5'): fname_h5 = fname else: return get_unparseable(fname) extract = { 'mldata': 0, 'name': '', 'comment': '', 'names': [], 'types': [], 'data': [], } try: h5 = h5py.File(fname, 'r') except: return extract attrs = ('mldata', 'name', 'comment') for a in attrs: if a in h5.attrs: extract[a] = h5.attrs[a] if '/data_descr/names' in h5: extract['names'] = h5['/data_descr/names'][:].tolist()[:NUM_EXTRACT] extract['names_cut'] = copy.copy(extract['names']) for i in xrange(len(extract['names_cut'])): if len(extract['names_cut'][i]) > LEN_EXTRACT: extract['names_cut'][i] = extract['names_cut'][i][:LEN_EXTRACT-2] + '...' if len(list(h5['/data_descr/names'])) > NUM_EXTRACT: extract['names_cut'].append('...') if '/data_descr/types' in h5: extract['types'] = h5['/data_descr/types'][:].tolist()[:NUM_EXTRACT] extract['data'] = _get_extract_data(h5) h5.close() return extract def is_binary(fname): """Return true if the given file is binary. @param fname: name of file to check if binary @type fname: string @return: if file is binary @rtype: boolean """ f = open(fname, 'rb') try: CHUNKSIZE = 1024 while 1: chunk = f.read(CHUNKSIZE) if '\0' in chunk: # found null byte f.close() return 1 if len(chunk) < CHUNKSIZE: break # done finally: f.close() return 0 def get_unparseable(fname): """Get data from unparseable files @param fname: filename to get data from @type fname: string @return: raw extract from unparseable file @rtype: dict with 'attribute' data """ if zipfile.is_zipfile(fname): intro = 'ZIP archive' f = zipfile.ZipFile(fname) data = ', '.join(f.namelist()) f.close() elif tarfile.is_tarfile(fname): intro = '(Zipped) TAR archive' f = tarfile.TarFile.open(fname) data = ', '.join(f.getnames()) f.close() else: intro = 'Format not HDF5, zip or tar archive. Will parse the following text:' if is_binary(fname): data = '' else: f = open(fname, 'r') i = 0 data = [] for l in f: data.append(l) i += 1 if i > NUM_EXTRACT: break f.close() data = "\n".join(data) return {'data': [[intro, data]]} def get_uncompressed(fname): """Get name of uncompressed Data file. The given file must be a zipfile/tarball and must contain exactly 1 file to be uncompressed. The returned name represents a file actually created in the file system which the user of this method might have to os.remove(). @param fname: (possibly) compressed filename @type fname: string @return: uncompressed filename if file is compressed, None otherwise @rtype: string """ src = None path = os.path.dirname(fname) # archives if zipfile.is_zipfile(fname): src = zipfile.ZipFile(fname) fnames = src.namelist() elif tarfile.is_tarfile(fname): src = tarfile.open(fname) fnames = src.getnames() if src: if len(fnames) == 1: src.extract(fnames[0], path) return os.path.join(path, fnames[0]) else: return None # gz/bz2 try: src = bz2.BZ2File(fname) except: try: src = gzip.open(fname, 'r') except: return None try: base = os.path.basename(fname) name = os.path.join(path, '.'.join(base.split('.')[:-1])) dst = open(name, 'w') dst.write(src.read()) dst.close() except: if os.path.exists(name) and not os.path.isdir(name): os.remove(name) name = None src.close() return name def _find_dset(fname, output_variables): """Find the dataset in given contents that contains the output_variable(s). This would be easy if all the data was just in one blob, but it may be in several datasets as defined by contents['ordering'], e.g. in contents['label'] or contents['data'] or contents['nameofvariable']. @param output_variables: index of output_variables to look for @type output_variables: integer @return: dataset corresponding to output_variables @rtype: list of integer """ h5 = h5py.File(fname, 'r') dset = None ov = output_variables for name in h5['/data_descr/ordering']: path = '/data/' + name path_indptr = path+'_indptr' path_indices = path+'_indices' if path_indptr in h5['/data'] and path_indices in h5['/data']: A = csc_matrix( (h5['/data/data'], h5['/data/indices'], h5['/data/indptr']) ) if ov == 0: for i in xrange(A.shape[0]): if ov == 0: dset = A[i].todense().tolist() break # inner loop else: ov -= 1 else: data = h5[path][...] if len(data.shape) == 1: # datasets with shape (x,) if ov == 0: dset = data else: ov -= 1 else: # datasets with shape (x,y) for i in xrange(len(data)): if ov == 0: dset = data[i].tolist() break # inner loop else: ov -= 1 if dset is not None: break # outer loop h5.close() return dset def get_attribute_types(fname): types=set() dt = h5py.special_dtype(vlen=str) try: h5 = h5py.File(fname, 'r') for o in h5['/data_descr/ordering']: indptr_name='/data/' + o + '_indptr' indices_name='/data/' + o + '_indices' if indptr_name in h5 and indices_name in h5: types += 'Sparse Matrix' else: try: types += h5['/data_descr/types'][o] except: t=h5['/data/'][o].dtype if t==dt: types.add("String") elif t in (numpy.int64, numpy.int32): types.add("Integer") elif t in (numpy.float64, numpy.float32): types.add("Floating Point") else: types.add(str(t)) h5.close() except: pass return ','.join(list(types)) def get_correct(fname, test_idx, output_variables): """Get correct results from given file from given example indices at given attribute index. Used for solving tasks. @param fname: name of Data file @type fname: string @param test_idx: example indices for the test @type test_idx list of integer @param output_variables: index of attribute to get values for @type output_variables: integer @return: correct results @rtype: list of integer """ dset = _find_dset(fname, output_variables) correct = [] for idx in test_idx: correct.append(dset[idx]) return correct mldata-utils-0.4.1/ml2h5/converter/0000700000175000017500000000000011477311160016067 5ustar sonnesonnemldata-utils-0.4.1/ml2h5/converter/h5_mat.py0000700000175000017500000000566711472443701017641 0ustar sonnesonneimport h5py, numpy from scipy.io import savemat, loadmat from basehandler import BaseHandler class H5_MAT(BaseHandler): """Handle Matlab files.""" def __init__(self, *args, **kwargs): super(H5_MAT, self).__init__(*args, **kwargs) def read(self): matf = loadmat(self.fname, squeeze_me=False, chars_as_strings=True, mat_dtype=True, struct_as_record=True) for k in ('__header__', '__globals__', '__version__'): if matf.has_key(k): del matf[k] for k in matf.keys(): if matf[k].dtype == numpy.object: # asume cell of strings cell = [] for i in matf[k][0]: if len(i[0]): cell.append(str(i[0])) else: cell.append('') matf[k]=cell elif matf[k].shape[0] == 1: # row vectors are expanded matf[k]=matf[k][0] data = matf if matf.has_key('mldata_descr_ordering'): ordering = matf['mldata_descr_ordering'] del(matf['mldata_descr_ordering']) else: ordering = matf.keys() def strip_type(x): if x.startswith('double'): return int(x[6:]) elif x.startswith('int'): return int(x[3:]) elif x.startswith('str'): return int(x[3:]) else: return x ordering.sort(cmp=lambda x,y: cmp(strip_type(x),strip_type(y))) return { 'name': self.get_name(), 'comment': 'matlab', 'names': [], 'ordering': ordering, 'data': data, } def write(self, data): group=self.get_data_group(data) d=data[group] for k in d.keys(): if type(d[k])==list and len(d[k])>0 and type(d[k][0])==str: cell = array([ numpy.array(unicode(i)) for i in d[k] ], dtype=numpy.object) d[k] = cell elif type(d[k])==numpy.ndarray and len(d[k])>0 and type(d[k][0])==str: cell = numpy.empty((1,len(d[k])), dtype=numpy.object) for i in xrange(len(d[k])): cell[0,i]=numpy.array(unicode(d[k][i]), dtype='U') d[k] = cell elif type(d[k])==numpy.ndarray and len(d[k])>0 and d[k][0].dtype == numpy.object: cell = numpy.empty((1,len(d[k])), dtype=numpy.object) for i in xrange(len(d[k])): cell[0,i]=numpy.array(unicode(d[k][i]), dtype='U') d[k] = cell d['mldata_descr_ordering']=numpy.array(data['ordering'],dtype=numpy.object) savemat(self.fname, d, appendmat=False, oned_as='row', format='5', long_field_names=True) mldata-utils-0.4.1/ml2h5/converter/h5_octave.py0000700000175000017500000003423311473007737020336 0ustar sonnesonneimport h5py, numpy from basehandler import BaseHandler from scipy.sparse import csc_matrix import ml2h5.converter class H5_OCTAVE(BaseHandler): """Handle Octave files.""" def __init__(self, *args, **kwargs): super(H5_OCTAVE, self).__init__(*args, **kwargs) def _check_header(self, octf): """evidence of octave conformity (disabled) @param octf: octave file @type octf: opened File @return: if header exists @rtype: boolean """ octf.seek(0) header=octf.readline(15) if header.startswith('# Created by '): return True else: return False def _next_attr(self, octf): """Returns the next atribute in the octave file @param octf: octave file @type octf: opened File @return (name,data): name and proper matrix of the atribute """ name='' data=[] meta = self._read_meta(octf) # throw conversion exceptions ? if meta==None: return None if not meta.has_key('name'): return None if not meta.has_key('dtype'): return None if meta['dtype']=='scalar': data=self._read_matrix(octf,1,1) elif meta['dtype']=='matrix': data=self._read_matrix(octf,meta['columns'],meta['rows']) elif meta['dtype']=='int32 matrix': data=self._read_matrix(octf,1,1,mtype='int') elif meta['dtype']=='sparse matrix': data=self._read_sparse_matrix(octf,meta['columns'],meta['rows']) elif meta['dtype']=='cell': data=self._read_cellarray(octf,meta['columns'],meta['rows']) elif meta['dtype']=='string': data=self._read_sq_string(octf,meta['length'],meta['elements']) elif meta['dtype']=='sq_string': data=self._read_sq_string(octf,meta['length'],meta['elements']) else: # unsupported data type. format exception ? return None return { 'name':meta['name'], 'data':data, } def _read_meta(self,octf): """Returns the meta information of the next atribute in the octave file @param octf: octave file @type octf: opened File @return meta: dictionary of attribute meta information """ meta={} line = octf.readline() if not line: return None while not line.startswith('#'): if line == '': return None line = octf.readline() # metadata while line and line.startswith('#'): sp=line.split(': ') if sp[0]=='# name': if meta.has_key('name'): break meta['name']=sp[1][:-1] if sp[0]=='# type': meta['dtype']=sp[1][:-1] if sp[0]=='# rows': meta['rows']=int(sp[1]) if sp[0]=='# columns': meta['columns']=int(sp[1]) if sp[0]=='# elements': meta['elements']=int(sp[1]) if sp[0]=='# length': meta['length']=int(sp[1]) if sp[0]=='# ndims': meta['ndims']=int(sp[1]) lpos=octf.tell() line = octf.readline() octf.seek(lpos) return meta def _read_sq_string(self,octf,len,ele): """Returns the next string sequence atribute in the octave file @param octf: octave file @type octf: opened File @return data: array of strings """ data='' data=octf.readline()[:-1] return data def _read_cellarray(self,octf,col,row): """Returns the next cellarray atribute in the octave file @param octf: octave file @type octf: opened File @return data: array """ data=[] for r in range(row): for c in range(col): meta=self._read_meta(octf) if not meta['dtype']=='sq_string': return None else: data.append(self._read_sq_string(octf,meta['length'],meta['elements'])) return data def _read_sparse_matrix(self,octf,col,row): """Returns the next sparse matrix in the octave file @param octf: octave file @type octf: opened File @return data: csc_matrix """ tmp_data=numpy.array(self._read_matrix(octf,0,0)).T data=csc_matrix((tmp_data[2],tmp_data[0:2]-1),shape=(row,col)) data.sort_indices() return data def _read_matrix(self,octf,col,row,mtype='float'): """Returns the data of a matrix atribute in the octave file @param octf: octave file @type octf: opened File @return data: matrix """ data=[] line=octf.readline() lpos=octf.tell() if mtype=='int': if line.startswith(' '): line=line[1:] sp=line[:-1].split(' ') row=int(sp[0]) col=int(sp[1]) line=octf.readline() while line and not line.startswith('#'): if line.startswith(' '): line=line[1:] sp=line[:-1].split(' ') if mtype=='int': conv_sp=[] try: for i in sp: conv_sp.append(int(i)) except ValueError: raise ConversionError('unexpected data type') if mtype=='float': conv_sp=[] try: for i in sp: conv_sp.append(float(i)) except ValueError: raise ConversionError('unexpected data type') data.append(conv_sp) lpos=octf.tell() line = octf.readline() octf.seek(lpos) out = numpy.array(data) # unknown matrix shape ? if not (row==0 and col==0): out.shape=(row,col) if out.shape[0]==1: out.shape=(out.shape[1],) return out def read(self): data={} names=[] octf = open(self.fname, 'r') # header check if not self._check_header(octf): raise ml2h5.converter.ConversionError('Header check failed') attr=self._next_attr(octf) while attr and attr['name']!='': if attr['name']!='__nargin__': data[attr['name']]=attr['data'] names.append(attr['name']) attr=self._next_attr(octf) if (data.keys==[]): raise ml2h5.converter.ConversionError('empty conversion') return { 'name': self.get_name(), 'comment': 'octave', 'names':[], 'ordering': names, 'data': data, } def _oct_header(self): return '# Created by mldata.org for Octave 3.0.1\n' def _num_matrix(self,m): if type(m)!= numpy.ndarray: if type(m)==list: try: m=numpy.array(m) except: return None else: return None if len(m.shape) == 1: out=[] for i in m: try: if int(i)!=i: raise ValueError out.append(int(i)) except: try: out.append(float(i)) except: return None elif len(m.shape) == 2: out=[] for i in m: row=[] for j in i: try: if int(j)!=j: raise ValueError row.append(int(j)) except: try: row.append(float(j)) except: return None out.append(row) else: return None return out def _print_meta(self,of,attr,name): """Return a string of metainformation @return meta: string of attr informations """ attr_num=self._num_matrix(attr) of.write('# name: ' + str(name) + '\n') if type(attr) == numpy.ndarray: if attr.shape ==(): of.write('# type: sq_string\n') of.write('# elements: 1\n') elif attr.shape ==(1,1) or attr.shape==(1,): of.write('# type: scalar\n') elif len(attr.shape)==1: if attr_num==None: of.write('# type: cell\n') of.write('# rows: 1\n') of.write('# columns: ' + str(len(attr)) + '\n') else: if attr.dtype in ['int32', 'int64']: of.write('# type: int32 matrix\n') of.write('# ndims 2\n') else: of.write('# type: matrix\n') of.write('# rows: 1\n') try: of.write('# columns: ' + str(attr.shape[0]) + '\n') except IndexError: of.write('# columns: 1\n') else: if attr.dtype in ['int32', 'int64']: of.write('# type: int32 matrix\n') of.write('# ndims 2\n') else: of.write('# type: matrix\n') try: of.write('# rows: ' + str(attr.shape[0]) + '\n') except IndexError: of.write('# rows: 1\n') try: of.write('# columns: ' + str(attr.shape[1]) + '\n') except IndexError: of.write('# columns: 1\n') elif type(attr)== csc_matrix: of.write('# type: sparse matrix\n') of.write('# nnz: '+str(attr.nnz) + '\n') of.write('# rows: '+str(attr.shape[0]) + '\n') of.write('# columns: '+str(attr.shape[1]) + '\n') elif type(attr) == list: of.write('# type: cell\n') of.write('# rows: 1\n') of.write('# columns: ' + str(len(attr)) + '\n') else: of.write('# type: sq_string\n') of.write('# elements: 1\n') return True def _print_data(self, of, attr): """Return a string of data @return data: string of attr content """ if attr==None: return False attr_num=self._num_matrix(attr) # matrix or scalar or cell array if type(attr) == numpy.ndarray: # sq_string if attr.shape==(): of.write('# length: ' + str(len(str(attr))) + '\n') of.write(str(attr) + '\n\n') # scalar elif attr.shape==(1,1): of.write(str(attr_num[0][0]) + '\n') elif attr.shape==(1,): of.write(str(attr_num[0]) + '\n') # matrix elif len(attr.shape)==2: # int32 matrix if attr.dtype in ['int32','int64']: of.write(' ' + str(attr.shape[0]) + ' ' + str(attr.shape[1]) + '\n') for i in attr: for j in i: of.write(' ' + str(j) + '\n') # float matrix else: for i in attr_num: for j in i: of.write(' ' + str(j)) of.write('\n') # matrix (vector) or cell array elif len(attr.shape)==1: # matrix if attr_num!=None: # int32 matrix if attr.dtype in ['int32', 'int64']: of.write(' 1 ' + str(attr.shape[0]) + '\n') for i in attr: of.write(' ' + str(i) + '\n') # float matrix else: for i in attr_num: of.write(' ' + str(i)) of.write('\n') else: # cell array for i in attr: of.write('# name: \n') of.write('# type: sq_string\n') of.write('# elements: 1\n') of.write('# length: ' + str(len(i)) + '\n') of.write(str(i) + '\n\n') # sparse matrix elif type(attr) == csc_matrix: dat=self._num_matrix(attr.data) indptr=attr.indptr indices=attr.indices for i in xrange(attr.shape[1]): out=[] for j in xrange(indptr[i],indptr[i+1]): if dat[j]==int: out.append("%d %d %d\n" % (indices[j]+1, i+1, dat[j])) else: out.append("%d %d %.15g\n" % (indices[j]+1, i+1, dat[j])) of.write(''.join(out)) # more clean but slower code #indices=attr.nonzero() #for i in range(len(attr.data)): # of.write(str(indices[0][i]+1) + ' ' + str(indices[1][i]+1) + ' ' + str(dat[i]) + '\n') # cell array elif type(attr) == list: for i in attr: of.write('# name: \n') of.write('# type: sq_string\n') of.write('# elements: 1\n') of.write('# length: ' + str(len(i)) + '\n') of.write(str(i) + '\n\n') # single string else: of.write('# length: ' + str(len(attr)) + '\n') of.write(str(attr) + '\n\n') return True def write(self, data): group=self.get_data_group(data) of = open(self.fname,'w') of.writelines(self._oct_header()) for o in data['ordering']: self._print_meta(of,data[group][o], o) self._print_data(of,data[group][o]) of.close() mldata-utils-0.4.1/ml2h5/converter/__init__.py0000600000175000017500000001633211475246572020223 0ustar sonnesonne""" Convert from and to HDF5 (spec of mldata.org) """ ALLOWED_SEPERATORS = (None, ',', ' ', '\t') TO_H5 = ['libsvm', 'arff', 'csv', 'matlab', 'octave'] FROM_H5 = ['libsvm','arff', 'csv', 'matlab', 'octave', 'xml'] EPSILON = 1e-15 #1 MB is maximum line len for autodetection AUTODETECTION_MAXBUFLEN=1*1024*1024 import os, sys, numpy, h5py import subprocess from gettext import gettext as _ from scipy.sparse import csc_matrix import ml2h5.fileformat from h5_arff import H5_ARFF from h5_libsvm import H5_LibSVM from h5_csv import H5_CSV from h5_mat import H5_MAT from h5_octave import H5_OCTAVE #from h5_uci import H5_UCI from basehandler import BaseHandler HANDLERS = { 'libsvm': H5_LibSVM, 'arff': H5_ARFF, 'csv': H5_CSV, 'matlab': H5_MAT, 'octave' : H5_OCTAVE, #'uci' : H5_UCI, 'h5': BaseHandler } class ConversionError(Exception): def __init__(self, value): self.value = value def __str__(self): return repr(self.value) def print_error(self): print self.value class Converter(object): """Convert and verify conversion. @ivar fname_in: name of in-file @type fname_in: string @ivar fname_out: name of out-file @type fname_out: string @ivar format_in: format of in-file @type format_in: string @ivar format_out: format of out-file @type format_out: string @ivar handler_in: handler object to read data in a specific format @type handler_in: derivate of BaseHandler @ivar handler_out: handler object to write data in a specific format @type handler_out: derivate of BaseHandler """ def __init__(self, fname_in, fname_out, format_in=None, format_out=None, seperator=None, attribute_names_first=False, merge=True, type='data'): """ @param fname_in: name of in-file @type fname_in: string @param fname_out: name of out-file @type fname_out: string @param format_in: format of in-file @type format_in: string @param format_out: format of out-file @type format_out: string @param seperator: seperator to seperate variables in examples @type seperator: string @param attribute_names_first: first line contains attributes (for CSV only) @type attribute_names_first: boolean """ self.fname_in = fname_in self.fname_out = fname_out self.type = type if format_in: self.format_in = format_in else: self.format_in = ml2h5.fileformat.get(fname_in) if format_out: self.format_out = format_out else: self.format_out = ml2h5.fileformat.get(fname_out) # handle h5 -> xml via h5dump if self.format_in == 'h5' and self.format_out == 'xml': return try: self.handler_in = HANDLERS[self.format_in](fname_in, seperator,merge=merge) if self.format_in == 'csv': self.handler_in.attribute_names_first = attribute_names_first self.handler_out = HANDLERS[self.format_out](fname_out, seperator,merge=merge) if self.format_out == 'csv': self.handler_out.attribute_names_first = attribute_names_first except KeyError: raise ConversionError( 'Unknown conversion pair %s to %s!' % (self.format_in, self.format_out)) except Exception, e: # reformat all other exceptions to ConversionError raise ConversionError, ConversionError(str(e)), sys.exc_info()[2] def read(self): try: return self.handler_in.read() except Exception, e: # reformat all exceptions to ConversionError raise ConversionError, ConversionError(str(e)), sys.exc_info()[2] def run(self, verify=False, remove_out=True): """Convert to/from HDF5. @param verify: verify if data in output is same as data in input @type verify: boolean @param remove_out: if output file shall be removed before running. @type remove_out: boolean """ # sometimes it seems files are not properly overwritten when opened by # 'w' during run(). if remove_out and os.path.exists(self.fname_out): os.remove(self.fname_out) try: if self.format_in == 'h5' and self.format_out == 'xml': cmd = 'h5dump --xml ' + self.fname_in + ' > ' + self.fname_out if not subprocess.call(cmd, shell=True) == 0: raise ConversionError, 'Failed conversion of %s to XML' % (self.fname_in) else: data = self.handler_in.read() self.handler_out.write(data) except Exception, e: # reformat all exceptions to ConversionError raise ConversionError, ConversionError(str(e)), sys.exc_info()[2] def _compare(self, A, B): """Compare given matrices A and B. Used by verification process. @param A: list A to compare @type A: list of list @param B: list B to compare @type B: list of list """ if type(A) == csc_matrix: try: if all(A.indices==B.indices) and all(A.indptr==B.indptr) and all(A.data==B.data): return True else: return False except: return False if type(A[0])==numpy.ndarray: xrange_A0 = xrange(len(A[0])) for i in xrange(len(A)): Ai = A[i] Bi = B[i] for j in xrange_A0: try: if abs(Ai[j] - Bi[j]) > EPSILON: return False except TypeError: #string if str(Ai[j]) != str(Bi[j]): return False else: if type(A) == numpy.ndarray and \ type(B) == numpy.ndarray and \ A.shape != B.shape: return False for j in xrange(len(A)): try: if abs(A[j] - B[j]) > EPSILON: return False except TypeError: #string if str(A[j]) != str(B[j]): return False return True def verify(self): """Verify that data in given files is the same. @return: true if verification succeeds @rtype: boolean @raises: ConversionError """ if self.format_in == 'uci': raise ConversionError('Cannot verify UCI data format, %s!' % self.fname_in) if self.format_out == 'uci': raise ConversionError('Cannot verify UCI data format, %s!' % self.fname_out) data_in = self.handler_in.read() data_out = self.handler_out.read() for i in xrange(len(data_in['ordering'])): name_in = data_in['ordering'][i] name_out = data_out['ordering'][i] if not self._compare(data_in['data'][name_in], data_out['data'][name_out]): raise ConversionError( 'Verification failed! Data of %s != %s ("%s" not matching "%s")' % (self.fname_in, self.fname_out, name_in, name_out) ) return True mldata-utils-0.4.1/ml2h5/converter/arff.py0000600000175000017500000002514611465067567017411 0ustar sonnesonne# -*- coding: utf-8 -*- # Copyright (c) 2008, Mikio L. Braun, Cheng Soon Ong, Soeren Sonnenburg # All rights reserved. # # Redistribution and use in source and binary forms, with or without # modification, are permitted provided that the following conditions are # met: # # * Redistributions of source code must retain the above copyright # notice, this list of conditions and the following disclaimer. # * Redistributions in binary form must reproduce the above # copyright notice, this list of conditions and the following disclaimer # in the documentation and/or other materials provided with the # distribution. # * Neither the names of the Technical University of Berlin, ETH # Zürich, or Fraunhofer FIRST nor the names of its contributors may be # used to endorse or promote products derived from this software without # specific prior written permission. # # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS # "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT # LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR # A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT # OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, # SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT # LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, # DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY # THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. import re, sys, numpy class ArffFile(object): """An ARFF File object describes a data set consisting of a number of data points made up of attributes. The whole data set is called a 'relation'. Supported attributes are: - 'numeric': floating point numbers - 'string': strings - 'nominal': taking one of a number of possible values Not all features of ARFF files are supported yet. The most notable exceptions are: - no sparse data - no support for date and relational attributes Also, parsing of strings might still be a bit brittle. You can either load or save from files, or write and parse from a string. You can also construct an empty ARFF file and then fill in your data by hand. To define attributes use the define_attribute method. Attributes are: - 'relation': name of the relation - 'attributes': names of the attributes - 'attribute_types': types of the attributes - 'attribute_data': additional data, for example for nominal attributes. - 'comment': the initial comment in the file. Typically contains some information on the data set. - 'data': the actual data, by data points. """ def __init__(self): """Construct an empty ARFF structure.""" self.relation = '' self.attributes = [] self.attribute_types = dict() self.attribute_data = dict() self.comment = [] self.data = [] pass @staticmethod def load(filename): """Load an ARFF File from a file.""" o = open(filename) s = o.read() a = ArffFile.parse(s) o.close() if not a.relation: raise Exception('Not an ARFF file: ' + filename) return a def _rm_ticks(self, item): """Remove ticks from item if it is a string. Some attributes are surrounded by unnecessary ticks. @param item: item to check for ticks @type item: any, preferably str @return: unmodified item or item with ticks removed @rtype: type(item) """ try: # a few attributes are designated strings by "'" if item[0] == "'" and item[-1] == "'": return item[1:-1] except TypeError: pass return item @staticmethod def parse(s): """Parse an ARFF File already loaded into a string.""" a = ArffFile() a.state = 'comment' a.lineno = 1 for l in s.splitlines(): a.__parseline(l) a.lineno += 1 return a def save(self, filename): """Save an arff structure to a file.""" o = open(filename, 'w') o.write(self.write()) o.close() def write(self): """Write an arff structure to a string.""" o = [] o.append('% ' + re.sub("\n", "\n% ", self.comment)) o.append("@relation " + self.esc(self.relation)) for a in self.attributes: at = self.attribute_types[a] if at == 'numeric': o.append("@attribute " + self.esc(a) + " numeric") elif at == 'string': o.append("@attribute " + self.esc(a) + " string") elif at == 'date': o.append("@attribute " + self.esc(a) + " date " + \ ''.join(self.attribute_data[a])) elif at == 'nominal': o.append("@attribute " + self.esc(a) + " {" + ','.join(self.attribute_data[a]) + "}") else: raise Exception("Type " + at + " not supported for writing!") o.append("\n@data") for d in self.data: line = [] for e, a in zip(d, self.attributes): at = self.attribute_types[a] if at == 'numeric': if type(e) in [numpy.int32,numpy.int64]: line.append(str(int(e))) else: line.append(str(float(e))) elif at == 'string': line.append(self.esc(e)) elif at == 'date': line.append(self.esc(e)) elif at == 'nominal': line.append(e) else: raise Exception("Type " + at + " not supported for writing!") o.append(','.join(map(str, line))) return "\n".join(o) + "\n" def esc(self, s): "Escape a string if it contains spaces" try: if re.match(r'\s', s): return "\'" + str(s) + "\'" except TypeError: pass return str(s) def define_attribute(self, name, atype, data=None): """Define a new attribute. atype has to be one of 'numeric', 'string', 'date' or 'nominal'. For nominal or date attributes, pass the possible values as data.""" n = self._rm_ticks(name) self.attributes.append(n) self.attribute_types[n] = atype self.attribute_data[n] = data def __parseline(self, l): if self.state == 'comment': if len(l) > 0 and l[0] == '%': self.comment.append(l[2:]) else: self.comment = '\n'.join(self.comment) self.state = 'in_header' self.__parseline(l) elif self.state == 'in_header': ll = l.lower().strip() if ll.startswith('@relation '): self.__parse_relation(l) if ll.startswith('@attribute '): self.__parse_attribute(l) if ll.startswith('@data'): self.state = 'data' elif self.state == 'data': if len(l) > 0 and l[0] != '%': self.__parse_data(l) def __parse_relation(self, l): l = l.split() self.relation = l[1] def __parse_attribute(self, l): p = re.compile(r'[a-zA-Z_][a-zA-Z0-9_/+-.*\(\)]*|\{[^\}]+\}|\'[^\']+\'|\"[^\"]+\"') l = [s.strip() for s in p.findall(l)] name = l[1] atype = l[2] atypel = atype.lower() if (atypel == 'real' or atypel == 'numeric' or atypel == 'integer'): self.define_attribute(name, 'numeric') elif atypel == 'string': self.define_attribute(name, 'string') elif atypel == 'date': self.define_attribute(name, 'date', l[3]) elif atype[0] == '{' and atype[-1] == '}': values = [s.strip () for s in atype[1:-1].split(',')] self.define_attribute(name, 'nominal', values) else: self.__print_warning("unsupported type " + atype + " for attribute " + name + ".") def __parse_data(self, line): l = [s.strip() for s in line.split(',')] if len(l) == 1: l = [s.strip() for s in line.split()] if not l[-1].strip(): # remove trailing empty item l.pop() if len(l) != len(self.attributes): self.__print_warning("contains wrong number of values") return datum = [] for n, v in zip(self.attributes, l): at = self.attribute_types[n] if at == 'numeric': if v == '?' or v == '': datum.append(numpy.nan) elif re.match(r'[+-]?[0-9]*(?:\.[0-9]*(?:[eE]-?[0-9]+)?)?', v): try: datum.append(int(v)) except ValueError: datum.append(float(v)) else: self.__print_warning('non-numeric value %s for numeric attribute %s' % (v, n)) return elif at == 'string': datum.append(self._rm_ticks(v)) elif at == 'date': datum.append(self._rm_ticks(v)) elif at == 'nominal': if v in self.attribute_data[n]: datum.append(v) elif v == '?': datum.append(numpy.nan) else: self.__print_warning('incorrect value %s for nomial attribute %s' % (v, n)) return self.data.append(datum) def __print_warning(self, msg): #print ('Warning (line %d): ' % self.lineno) + msg pass def dump(self): """Print an overview of the ARFF file.""" print "Relation " + self.relation print " With attributes" for n in self.attributes: if self.attribute_types[n] != 'nominal': print " %s of type %s" % (n, self.attribute_types[n]) else: print (" " + n + " of type nominal with values " + ', '.join(self.attribute_data[n])) for d in self.data: print d if __name__ == '__main__': if False: a = ArffFile.parse("""% yes % this is great @relation foobar @attribute foo {a,b,c} @attribute bar real @data a, 1 b, 2 c, d d, 3 """) a.dump() a = ArffFile.load('../examples/diabetes.arff') print a.write() mldata-utils-0.4.1/ml2h5/converter/basehandler.py0000600000175000017500000002547511474776561020751 0ustar sonnesonneimport os, h5py, numpy from scipy.sparse import csc_matrix import ml2h5.task from ml2h5 import VERSION_MLDATA from ml2h5.converter import ALLOWED_SEPERATORS class BaseHandler(object): """Base handler class. It is the base for classes to handle different data formats. It implicitely handles HDF5. @cvar str_type: string type to be used for variable length strings in h5py @type str_type: numpy.dtype @ivar fname: name of file to handle @type fname: string @ivar seperator: seperator to seperate variables in examples @type seperator: string """ str_type = h5py.new_vlen(numpy.str) def __init__(self, fname, seperator=None, compression=None, merge=False): """ @param fname: name of in-file @type fname: string @param seperator: seperator used to seperate examples @type seperator: string """ self.fname = fname self.compression = compression self.set_seperator(seperator) self.merge = merge def set_seperator(self, seperator): """Set the seperator to seperate variables in examples. @param seperator: seperator to use @type seperator: string """ if seperator in ALLOWED_SEPERATORS: self.seperator = seperator else: raise AttributeError(_("Seperator '%s' not allowed!" % seperator)) def warn(self, msg): """Print a warning message. @param msg: message to print @type msg: string """ return print 'WARNING: ' + msg def _convert_to_ndarray(self,path,val): """converts a attribut to a set of ndarrays depending on the datatype @param path: path of the attribute in the h5 file @type path: string @param val: data of the attribute @type val: csc_matrix/ndarray @rtype: list of (string,ndarray) tuples """ A=val out=[] dt = h5py.special_dtype(vlen=str) if type(A)==csc_matrix: # sparse out.append((path+'_indices', A.indices)) out.append((path+'_indptr', A.indptr)) out.append((path, A.data)) elif type(A)==list and len(A)>0 and type(A[0])==str: out.append((path, numpy.array(A, dtype=dt))) else: # dense out.append((path, numpy.array(A))) return out def get_data_as_list(self,data): """ this needs to `transpose' the data """ dl=[] group=self.get_data_group(data) lengths=dict() for o in data['ordering']: x=data[group][o] #if numpy.issubdtype(x.dtype, numpy.int): # data[group][o]=x.astype(numpy.float64) try: lengths[o]=data[group][o].shape[1] except (AttributeError, IndexError): lengths[o]=len(data[group][o]) l=set(lengths.values()) assert(len(l)==1) l=l.pop() for i in xrange(l): line=[] for o in data['ordering']: try: line.extend(data[group][o][:,i]) except: line.append(data[group][o][i]) dl.append(line) return dl def get_name(self): """Get dataset name from non-HDF5 file @return: comment @rtype: string """ # without str() it might barf return str(os.path.basename(self.fname).split('.')[0]) def get_data_group(self, data): if data and data.has_key('group'): return data['group'] else: return 'data' def get_descr_group(self, data): if data and data.has_key('group'): return data['group'] + '_descr' else: return 'data_descr' def get_datatype(self, values): """Get data type of given values. @param values: list of values to check @type values: list @return: data type to use for conversion @rtype: numpy.int32/numpy.double/self.str_type """ dtype = None for v in values: if isinstance(v, int): dtype = numpy.int32 elif isinstance(v, float): dtype = numpy.double else: # maybe int/double in string try: tmp = int(v) if not dtype: # a previous nan might set it to double dtype = numpy.int32 except ValueError: try: tmp = float(v) dtype = numpy.double except ValueError: return self.str_type return dtype def read(self): """Get data and description in-memory Retrieve contents from file. @return: example names, ordering and the examples @rtype: dict of: list of names, list of ordering and dict of examples """ # we want the exception handled elsewhere h5 = h5py.File(self.fname, 'r') contents = { 'name': h5.attrs['name'], 'comment': h5.attrs['comment'], 'mldata': h5.attrs['mldata'], } if contents['comment']=='Task file': contents['task']=dict() contents['ordering']=ml2h5.task.task_data_fields group='task' else: contents['data']=dict() contents['ordering']=h5['/data_descr/ordering'][...].tolist() group='data' contents['group']=group if '/%s_descr/names' % group in h5: contents['names']=h5['/%s_descr/names' % group][...].tolist() if '/%s_descr/types' % group in h5: contents['types'] = h5['/%s_descr/types' % group ][...] for name in contents['ordering']: vname='/%s/%s' % (group, name) sp_indices=vname+'_indices' sp_indptr=vname+'_indptr' if sp_indices in h5['/%s' % group] and sp_indptr in h5['/%s' % group]: contents[group][name] = csc_matrix((h5[vname], h5[sp_indices], h5[sp_indptr]) ) else: d = numpy.array(h5[vname],order='F') try: d=d['vlen'] except: pass contents[group][name] = d h5.close() return contents def read_data_as_array(self): """Read data from file, and return an array @return: an array with all data @rtype: numpy ndarray """ contents = self.read() #group = self.get_data_group(data) data = contents['data'] ordering = contents['ordering'] if len(data[ordering[0]].shape)>1: num_examples = data[ordering[0]].shape[1] else: num_examples = len(data[ordering[0]]) data_array = numpy.zeros((0, num_examples)) for cur_feat in ordering: data_array = numpy.vstack([data_array, data[cur_feat]]) return data_array.T def _get_merged(self, data): """Merge given data where appropriate. String arrays are not merged, but all int and all double are merged into one matrix. @param data: data structure as returned by read() @type data: dict @return: merged data structure @rtype: dict """ merged = {} ordering = [] path = '' idx = 0 merging = None group = self.get_data_group(data) for name in data['ordering']: val = data[group][name] if type(val) == csc_matrix: merging = None path = name merged[path] = val ordering.append(path) continue if name.endswith('_indices') or name.endswith('_indptr'): merging = None path = name merged[path] = val continue if len(val) < 1: continue t = type(val[0]) if t in [numpy.int32, numpy.int64]: if merging == 'int': merged[path].append(val) else: merging = 'int' path = 'int' + str(idx) ordering.append(path) merged[path] = [val] idx += 1 elif t == numpy.double: if merging == 'double': merged[path].append(val) else: merging = 'double' path = 'double' + str(idx) ordering.append(path) merged[path] = [val] idx += 1 else: # string or matrix merging = None if name.find('/') != -1: # / sep belongs to hdf5 path path = name.replace('/', '+') data['ordering'][data['ordering'].index(name)] = path else: path = name ordering.append(path) merged[path] = val data[group] = {} for k in merged: if len(merged[k])==1: merged[k] = merged[k][0] data[group][k] = numpy.array(merged[k]) data['ordering'] = ordering return data def write(self, data): """Write given data to HDF5 file. @param data: data to write to HDF5 file. @type data: dict of lists """ # we want the exception handled elsewhere h5 = h5py.File(self.fname, 'w') h5.attrs['name'] = data['name'] h5.attrs['mldata'] = VERSION_MLDATA h5.attrs['comment'] = data['comment'] data_group = self.get_data_group(data) descr_group = self.get_descr_group(data) try: group = h5.create_group('/%s' % data_group) for path, val in data[data_group].iteritems(): for path, val in self._convert_to_ndarray(path,val): group.create_dataset(path, data=val, compression=self.compression) group = h5.create_group('/%s' % descr_group) names = numpy.array(data['names']).astype(self.str_type) if names.size > 0: # simple 'if names' throws exception if array group.create_dataset('names', data=names, compression=self.compression) ordering = numpy.array(data['ordering']).astype(self.str_type) if ordering.size > 0: group.create_dataset('ordering', data=ordering, compression=self.compression) if 'types' in data: types = numpy.array(data['types']).astype(self.str_type) group.create_dataset('types', data=types, compression=self.compression) except: # just do some clean-up h5.close() os.remove(self.fname) raise else: h5.close() mldata-utils-0.4.1/ml2h5/converter/h5_arff.py0000700000175000017500000001003611472445351017763 0ustar sonnesonneimport h5py, numpy, copy import arff from basehandler import BaseHandler from scipy.sparse import csc_matrix import ml2h5.converter class H5_ARFF(BaseHandler): """Handle ARFF files. It uses the module arff provided by the dataformat project: http://mloss.org/software/view/163/ """ def __init__(self, *args, **kwargs): super(H5_ARFF, self).__init__(*args, **kwargs) def _get_types(self, af): """Get types from Arff file @param af: opened ARFF file @type af: arff.ArffFile @return: types in arff file @rtype: numpy.array """ # need to get it in the right order types = [] for name in af.attributes: t = af.attribute_types[name] if af.attribute_data[name]: t += ':' + ','.join(af.attribute_data[name]) types.append(t) return numpy.array(types) def read(self): data = {} names = [] ordering = [] af = arff.ArffFile.load(self.fname) for name in af.attributes: names.append(name) data[name] = [] for item in af.data: for i in xrange(len(data)): data[names[i]].append(item[i]) # conversion to proper data types for name, values in data.iteritems(): if af.attribute_types[name] == 'date': t = self.str_type else: t = self.get_datatype(values) data[name] = numpy.array(values).astype(t) ddict= { 'name': af.relation, 'comment': af.comment, 'types': self._get_types(af), 'names':names, 'ordering':copy.copy(names), 'data':data, } if self.merge == True: ddict = self._get_merged(ddict) return ddict def check_sparse(self, data): for k in data.keys(): d=data[k] if type(d)==csc_matrix: raise ml2h5.converter.ConversionError("Sparse matrices are not supported in ARFF files") def write(self, data): group=self.get_data_group(data) d=data[group] self.check_sparse(d) af = arff.ArffFile() af.data = self.get_data_as_list(data) if data.has_key('names') and len(data['names']): af.attributes = data['names'] else: attrs=[] idx=0 for o in data['ordering']: try: if len(d[o].shape)==1: num=1 else: num=d[o].shape[0] if d[o].dtype == numpy.double: prefix='double' elif d[o].dtype in [numpy.int32, numpy.int64 ]: prefix='int' else: prefix='str' except AttributeError: # list - so assume list of strings num=1 prefix='str' for i in xrange(num): attrs.append('%s%d' % (prefix,idx)) idx+=1 if len(attrs)==len(data['ordering']): attrs=data['ordering'] af.attributes = attrs af.relation = data['name'] af.comment = data['comment'] # handle arff types if 'types' in data: types = data['types'] else: types = [] for name in af.attributes: if name.startswith('int') or name.startswith('double'): types.append('numeric') elif name.startswith('date'): types.append('date') else: types.append('string') for i in xrange(len(types)): t = types[i].split(':') af.attribute_types[af.attributes[i]] = t[0] if len(t) == 1: af.attribute_data[af.attributes[i]] = None else: af.attribute_data[af.attributes[i]] = t[1].split(',') af.save(self.fname) mldata-utils-0.4.1/ml2h5/converter/h5_uci.py0000700000175000017500000001011111460302625017610 0ustar sonnesonneimport os, numpy from basehandler import BaseHandler class H5_UCI(BaseHandler): """Handle UCI files.""" def _get_comment(self): try: name = '.'.join(self.fname.split('.')[:-1]) if not name: name = '.'.join(self.fname.split('-')[:-1]) if os.path.exists(name + '.names'): fp = open(name + '.names', 'r') elif os.path.exists(name + '.info'): fp = open(name + '.info', 'r') else: return '' comment = ''.join(fp.readlines()) fp.close() return comment except ValueError: return '' except IOError: return '' def _ignore_line(self, line): line = line.strip() if not line: return True if line.startswith(';;;'): return True return False def _split(self, line): line=line.strip() items = line.split() for i in xrange(len(items)): if '\t' in items[i]: old=items.pop(i) new=old.split() for it in new[::-1]: items.insert(i, it) marker_start=line.find('"') if marker_start == -1: return items marker_start=line.find('"') marker_stop=line.find('"', marker_start+1) conc_item_idx=line conc_start=None for i in xrange(len(items)): if marker_start == -1 or marker_stop == -1: return items if conc_start and '"' in items[i]: for j in xrange(conc_start, i+1): items.pop(conc_start) conc_str=line[marker_start+1:marker_stop] marker_start=line.find('"', marker_stop+1) marker_stop=line.find('"', marker_start+1) items.insert(conc_start, conc_str) conc_start = None continue elif not conc_start and '"' in items[i]: conc_start = i return items def _parse(self): fp = open(self.fname, 'r') lineno = 0 num_items = None data = [] for line in fp: lineno += 1 if self._ignore_line(line): continue items = self._split(line) if not num_items: # do some init num_items = len(items) for i in xrange(num_items): data.append([]) for i in xrange(len(items)): item = items[i].strip() if not item: continue try: if item == '?': # missing value item = numpy.nan data[i].append(item) except IndexError: self.warn('Index Error in line ' + str(lineno) + ', column ' + str(i)) fp.close() return data def read(self): ordering = [] predata = self._parse() for i in xrange(len(predata)): arr = numpy.array(predata[i]) # everything is nan -> keep as str try: all_is_nan = numpy.isnan(arr).all() except AttributeError: all_is_nan = False if all_is_nan: arr = arr.astype(self.str_type) name = 'str' + str(i) else: try: arr = arr.astype(numpy.int) name = 'int' + str(i) except ValueError: try: arr = arr.astype(numpy.double) name = 'double' + str(i) except ValueError: # fallback to str arr = arr.astype(self.str_type) name = 'str' + str(i) data[name] = arr ordering.append(name) return { 'name': self.get_name(), 'comment': self._get_comment(), 'ordering':ordering, 'names':ordering, 'data':data, } mldata-utils-0.4.1/ml2h5/converter/h5_csv.py0000700000175000017500000000654311464536435017655 0ustar sonnesonneimport numpy, h5py, os, copy, numpy from scipy.sparse import csc_matrix from basehandler import BaseHandler import ml2h5.converter COMMENT = '# ' SEPERATOR = ',' class H5_CSV(BaseHandler): """Handle CSV files. @ivar attribute_names_first: first line contains attributes @type attribute_names_first: boolean """ def __init__(self, *args, **kwargs): super(H5_CSV, self).__init__(*args, **kwargs) self.attribute_names_first = False self.seperator = SEPERATOR # maybe more flexible in the future def _find_nan(self, value): """Find NaN values and make them proper numpy.nan. Used by map function in get_contents() @param value: value to check if it is NaN @type value: string """ if value == '?': return numpy.nan else: # str() important! when run thru webserver it will be unicode # otherwise and somehow results somewhere in nan being converted # to string 'na'... return str(value) def read(self): """ """ data = {} names = [] ordering = [] infile = open(self.fname, 'r') parsed = [] anf = self.attribute_names_first # copy value for line in infile: l = line.strip().split(self.seperator) if anf: names = l anf = False else: l = map(self._find_nan, l) parsed.append(l) infile.close() A = numpy.matrix(parsed).T for i in xrange(A.shape[0]): items = A[i].tolist()[0] t = self.get_datatype(items) if t == numpy.int32: name = 'int' + str(i) elif t == numpy.double: name = 'double' + str(i) else: name = 'str' + str(i) data[name] = numpy.array(items).astype(t) ordering.append(name) if not names: names = copy.copy(ordering) ddict = { 'name': self.get_name(), 'comment': 'CSV', 'names': names, 'ordering': ordering, 'data': data, } if self.merge == True: ddict = self._get_merged(ddict) return ddict def write(self, data): lengths=dict() for o in data['ordering']: try: lengths[o]=data['data'][o].shape[1] except (AttributeError, IndexError): lengths[o]=len(data['data'][o]) l=set(lengths.values()) assert(len(l)==1) l=l.pop() csv = open(self.fname, 'w') for i in xrange(l): line=[] for o in data['ordering']: d=data['data'][o] if type(d)==csc_matrix: raise ml2h5.converter.ConversionError("Sparse matrices are not supported in CSV files") try: line.extend(map(str, d[:,i])) except: try: if len(d[i]): line.append(str(d[i])) else: line.append('') except TypeError: line.append(str(d[i])) csv.write(self.seperator.join(line) + "\n") csv.close() return True mldata-utils-0.4.1/ml2h5/converter/h5_libsvm.py0000700000175000017500000001161611463271542020345 0ustar sonnesonneimport numpy, h5py, os, copy from scipy.sparse import csc_matrix from basehandler import BaseHandler import ml2h5.converter class H5_LibSVM(BaseHandler): """Handle LibSVM files.""" def __init__(self, *args, **kwargs): """Constructor. @ivar is_multilabel: if data is of type multilabel @type is_multilabel: boolean """ super(H5_LibSVM, self).__init__(*args, **kwargs) self.is_multilabel = False def convert_sparse(self, spmatrix): assert(type(spmatrix==csc_matrix)) A=spmatrix if A.nnz/numpy.double(A.shape[0]*A.shape[1]) < 0.5: # sparse return spmatrix return numpy.array(A.todense()) def get_comment(self): return 'LibSVM' def read(self): """Retrieves a SciPy Compressed Sparse Column matrix and labels from file. @return: compressed sparse column matrix + labels @rtype: list of scipy.sparse.csc_matrix and label tuple/2-d tuple (multilabel) """ indices_var = [] indptr_var = [0] ptr_var = 0 data_var = [] indices_lab = [] indptr_lab = [0] ptr_lab = 0 data_lab = [] label = [] for line in file(self.fname).xreadlines(): items = line.split() lab=items[0] dat=items[1:] lab=[int(float(i)) for i in lab.split(',')] if len(lab)>1: self.is_multilabel = True if self.is_multilabel: for idx in lab: indices_lab.append(idx) data_lab.append(1.) ptr_lab += 1 indptr_lab.append(ptr_lab) else: # only single label -> value is actual value label.append(lab[0]) for d in dat: v=d.split(':') indices_var.append(int(v[0]) - 1) data_var.append(numpy.double(v[1])) ptr_var += 1 indptr_var.append(ptr_var) if self.is_multilabel: data_lab=numpy.array(data_lab) indices_lab=numpy.array(indices_lab) indptr_lab=numpy.array(indptr_lab) label = csc_matrix( (data_lab, indices_lab, indptr_lab) ) label.sort_indices() else: label = numpy.array(label) data = csc_matrix( (numpy.array(data_var), numpy.array(indices_var), numpy.array(indptr_var)) ) data = self.convert_sparse(data) return { 'name': self.get_name(), 'comment': 'LibSVM', 'ordering': ['label', 'data'], 'names': [], 'data': { 'label' : label, 'data' : data} } def write(self, data): ordering=('label','data') if not set(data['data'].keys()).issubset(set(ordering)): raise ml2h5.converter.ConversionError('libsvm format needs data or label') t = type(data['data']['data']) if t not in (csc_matrix, numpy.ndarray): raise ml2h5.converter.ConversionError('libsvm format needs csc or dense matrix "data"') libsvm = open(self.fname, 'w') if 'label' in data['data'].keys(): if type(data['data']['label']) == csc_matrix: self.is_multilabel = True else: self.is_multilabel = False lengths=dict() for o in ordering: try: lengths[o]=data['data'][o].shape[1] except (AttributeError, IndexError): lengths[o]=len(data['data'][o]) l=set(lengths.values()) assert(len(l)==1) l=l.pop() for i in xrange(l): out = [] for o in ordering: d=data['data'][o] if o == 'label': if self.is_multilabel: labels = [] indptr=d.indptr indices=d.indices dat=d.data for j in xrange(indptr[i],indptr[i+1]): labels.append(str(indices[j])) out.append(','.join(labels)) else: out.append(str(d[i])) else: #data if type(d)==csc_matrix: indptr=d.indptr indices=d.indices dat=d.data for j in xrange(indptr[i],indptr[i+1]): if dat[j]==int: out.append('%d:%d' % (indices[j]+1,dat[j])) else: out.append('%d:%.15g' % (indices[j]+1,dat[j])) else: # dense for j in xrange(len(d)): out.append(str(j+1) + ':' + str(d[j,i])) libsvm.write(" ".join(out) + "\n") libsvm.close() return True mldata-utils-0.4.1/ml2h5/task.py0000600000175000017500000002253611474011477015413 0ustar sonnesonne""" Handle Task objects and files. This module heavily relies on the functionality required for http://mldata.org """ import os, h5py, numpy import ml2h5.data from . import VERSION_MLDATA from indexsplit import reduce_split_str COMPRESSION = None task_data_fields = ['train_idx', 'test_idx', 'input_variables', 'output_variables'] task_descr_fields = ['performance_measure','type'] def update_object(h5, name, value): if name in h5.keys(): del h5[name] h5[name] = value def get_splitnames(fnames): """Helper function to get names of splits. Get a name like test_idx, train_idx from given filenames. @param fnames: filenames to get splitnames from @type fnames: list of string @return: names of splits @rtype: list of strings """ names = [] for name in fnames: n = name.split(os.sep)[-1] if n.find('train') != -1 or n.find('.tr') != -1: names.append('train_idx') elif n.find('.val') != -1: names.append('validation_idx') elif n.find('test') != -1 or n.find('.t') != -1 or n.find('.r') != -1: names.append('test_idx') else: names.append(0) # replace unknown name by test if train exists or train if test exists if 0 in names: if 'train_idx' in names: names[names.index(0)] = 'test_idx' elif 'test_idx' in names: names[names.index(0)] = 'train_idx' return names def get_splitdata(fnames): """Get split data. @param fnames: filenames of related data files @type fnames: list of strings """ names = get_splitnames(fnames) data = {} offset = 0 for i in xrange(len(fnames)): count = sum(1 for line in open(fnames[i])) if names[i] in data: # in case we have multiple train/test idx data[names[i]].extend(range(offset, offset+count)) else: data[names[i]] = range(offset, offset+count) offset += count return data def add_data(fname, splitnames=None, variables=None): """Update a Task file by given splitfiles and input/output variables @param fname: name of the Task file @type fname: string @param splitnames: names of files to contain split data @type splitnames: list of strings @param variables: 'input': list with indices of the variables to be used to predict the target(s), 'output': index of the variable to be predicted from the features @type variables: dict of list of integer and integer @return: if Task file could be updated @rtype: boolean """ try: h5 = h5py.File(fname, 'a') except: return False if not 'task' in h5: group = h5.create_group('task') else: group = h5['task'] if variables: if 'input' in variables: group.create_dataset('input_variables', data=variables['input'], compression=COMPRESSION) if 'output' in variables: group.create_dataset('output_variables', data=variables['output'], compression=COMPRESSION) if splitnames: data = get_splitdata(splitnames) for k,v in data.iteritems(): if v: group.create_dataset(k, data=v, compression=COMPRESSION) h5.close() return True def _encode(text): """Encode given (utf-8) text to something h5py can digest. A bit annoying that this kind of stuff is necessary. @param text: text @type text: string @return: encoded string @rtype: string """ try: return text.encode('ascii', 'ignore') except AttributeError: return str(text) def update_description(h5, task): """Update description group in Task file. @param h5: opened HDF5 file @type h5: h5py.File @param task: Task object as of mldata.org @type task: repository.Task @return: if update was successful @rtype: boolean """ if not 'task_descr' in h5: group = h5.create_group('task_descr') else: group = h5['task_descr'] update_object(group, 'pub_date', _encode(task.pub_date)) update_object(group, 'version', task.version) update_object(group, 'slug', _encode(task.slug.text)) update_object(group, 'summary', _encode(task.summary)) update_object(group, 'description', _encode(task.description)) update_object(group, 'urls', _encode(task.urls)) # Without saving task, it does not have a primary key # and the following line complains task.save() update_object(group, 'publications',\ ''.join([_encode(p.title) for p in task.publications.all()])) update_object(group, 'input', _encode(task.input)) update_object(group, 'output', _encode(task.output)) update_object(group, 'performance_measure', _encode(task.performance_measure)) update_object(group, 'type', _encode(task.type)) update_object(group, 'data', _encode(task.data.name)) if task.data_heldback: update_object(group, 'data_heldback', _encode(task.data_heldback.name)) update_object(group, 'license', _encode(task.license.name)) update_object(group, 'tags', _encode(task.tags)) return True def update_data(h5, taskinfo=None): """Update data group in Task file. @param h5: opened HDF5 file @type h5: h5py.File @param taskinfo: data to write to Task file @type taskinfo: dict with indices train_idx, test_idx, input_variables, output_variables @return: if update was successful @rtype: boolean """ if not 'task' in h5: group = h5.create_group('task') else: group = h5['task'] if not taskinfo: return True for name in taskinfo: if taskinfo[name] is not None: if name in group: del group[name] group.create_dataset(name, data=taskinfo[name]) return True def get_taskinfo(fname): taskinfo = None format = ml2h5.fileformat.get(fname) if not format in ('matlab','h5','octave'): raise ml2h5.converter.ConversionError, 'Format not supported (only matlab, \ octave, h5) are supported' try: c = ml2h5.converter.Converter(fname, '/tmp/dummy_does_not_exist.h5', format_in=format, format_out='h5', attribute_names_first=False, merge=False, type='data') data = c.read() for g in ('data','task'): for f in task_data_fields: if data.has_key(g) and data[g].has_key(f): if not taskinfo: taskinfo=dict() taskinfo[f]=data[g][f] except ml2h5.converter.ConversionError: pass return taskinfo def update_or_create(fname, task, taskinfo=None): """Update or create Task file with data from given Task object. @param fname: full path of Task filename @type fname: string @param task: Task object as of mldata.org @type task: repository.Task @param taskinfo: data to write to Task file @type taskinfo: dict with indices train_idx, test_idx, input_variables, output_variables @return: if file could be updated / created @rtype: boolean """ try: if not os.path.exists(fname): h5 = h5py.File(fname, 'w') else: h5 = h5py.File(fname, 'a') except: return False h5.attrs['name'] = _encode(task.name) h5.attrs['mldata'] = VERSION_MLDATA h5.attrs['comment'] = 'Task file' error = False if not update_description(h5, task): error = True if not update_data(h5, taskinfo): error = True h5.close() return not error def get_extract(fname): """Get extract of Task file. @param fname: name of Task file @type fname: string @return: datasets from Task file @rtype: dict of lists """ extract = {} try: h5 = h5py.File(fname, 'r') except: return extract for t in task_data_fields: try: extract[t]=reduce_split_str(h5['task'][t][...]) except KeyError: pass for t in task_descr_fields: try: extract[t]=h5['task_descr'][t][...] except KeyError: pass h5.close() return extract def get_split_image(fname): extract = {} try: h5 = h5py.File(fname, 'r') except: return extract dsets = ['train_idx', 'test_idx'] for dset in dsets: path = '/task/' + dset if path in h5: extract[dset] = h5[path][...] h5.close() if len(extract['train_idx'].shape)==1: dim=1 else: dim=extract['train_idx'].shape[0] image_data=numpy.zeros([max(extract['train_idx'][-1],extract['test_idx'][-1])+1,dim]) image_data[extract['train_idx']]=1 image_data[extract['test_idx']]=2 return image_data.T def get_variables(fname): """Get input/output variables from given Data file. @param fname: name of Data file to retrieve variables from @type fname: string @return: input and output variables @rtype: dict of lists 'input' and 'output' """ # FIXME: this might be completely wrong bucket, num_attr = ml2h5.data.get_num_instattr(fname) return { 'input': range(num_attr)[1:], 'output': 0, } def get_test_output(fname): """Get test_idx and output_variables from given Task file.""" h5 = h5py.File(fname, 'r') test_idx = h5['/task/test_idx'][:] output_variables = h5['/task/output_variables'][...] h5.close() return test_idx, output_variables mldata-utils-0.4.1/README0000600000175000017500000000024011450541742014011 0ustar sonnesonneThis package supplements http://mldata.org and should give you all tools necessary to work with Data, Task and Solution. See INSTALL for installation details. mldata-utils-0.4.1/INSTALL0000600000175000017500000000004511450541742014165 0ustar sonnesonneTo install: python setup.py install mldata-utils-0.4.1/CHANGELOG0000600000175000017500000000462111470304160014343 0ustar sonnesonne0.4.0 2010/11/12 by Cheng Soon Ong * Added clientapi to interact with mldata.org via python on a client. 0.3.6 2010/08/27 by Sebastian Henschel * Added a fix when data.get_correct internally receives an array of array with values instead an array with values. * Added support for sparse matrices in data.get_correct. 0.3.5 2010/08/25 by Sebastian Henschel * Introduced task.get_test_output to get test_idx and output_variables from Task file. * Introduced data.get_correct() to get the 'correct' results from Data file. * Fixed minor issus when converting to octave/matlab. 0.3.4 2010/08/24 by Sebastian Henschel * Fixed an issue with data extracts. * Fixed an issue when updating Task files. 0.3.3 2010/08/23 by Sebastian Henschel * Fixed a few issues when converting to arff/octave/matlab. 0.3.2 2010/08/21 by Sebastian Henschel * task.create now includes handling of input/output variables. * Fixed a little error handling octave files. 0.3.1 2010/08/20 by Sebastian Henschel * Had removed too much from data.get_extract and put it back in. * Added safeguard for illegal task files with no output_variables. 0.3.0 2010/08/19 by Sebastian Henschel * Restructured package into more different modules. * Revamped conversion structure. * Bugfix re Task vs output variables. 0.2.4 2010/08/17 by Sebastian Henschel * Caught a few more error conditions when handlings Task. * Temporarily removed author from package information because it threw ugly error message on older python installations. 0.2.3 2010/08/17 by Sebastian Henschel * Removed label_dims and improved support for input/output variables for Tasks. * Created new module 'data' for better encapsulation. 0.2.2 2010/08/16 by Sebastian Henschel * Added extract function (and script) for Task datasets. * Moved extract function for Data from website to this tool. 0.2.1 2010/08/02 by Sebastian Henschel * Improved handling of Task files. 0.2.0 2010/07/21 by Sebastian Henschel * Introduced CHANGELOG. * Moved Task-related code into its own-module. * Added code to save data from Task object into Task file. 0.1.0 2010/07/13 by Sebastian Henschel * Initial release as package. mldata-utils-0.4.1/mleval/0000700000175000017500000000000011477311160014411 5ustar sonnesonnemldata-utils-0.4.1/mleval/util.py0000600000175000017500000000144011451374163015745 0ustar sonnesonne# Some helper functions follow def register(dic, name, fun): dic[name]=(fun, fun.__doc__) def compute_area(a,b): """ Helper function to compute area """ return 0.5*numpy.sum(numpy.abs(a[:-1]-a[1:])*numpy.abs(b[:-1]+b[1:])) def alternative_calcprcscore(out, lab): """ For sanity checking only: auRPC compute function from PASCAL Image challenge. code translated from matlab code from image challenge """ idx = numpy.argsort(-out) tp=lab[idx]>0 fp=lab[idx]<0 fp=1.0*numpy.cumsum(fp) tp=1.0*numpy.cumsum(tp) rec=tp/numpy.sum(lab>0) prec=tp/(fp+tp) ap=0 for t in numpy.arange(0,1.01, 0.01): el=prec[rec>=t] if len(el)>0: p=numpy.max(el) else: p=0 ap=ap+p/101.0 return ap mldata-utils-0.4.1/mleval/classification.py0000600000175000017500000001057311453025065017766 0ustar sonnesonneimport numpy from util import register pm = dict() def calcbal(out, lab): """ Computes the Balanced Error. Expects labels to be +/-1 and predictions sign(output) """ pidx=(lab==+1) perr=numpy.mean(out[pidx]>0) nidx=(lab==-1) nerr=numpy.mean(out[nidx]<0) return 0.5 * (perr+nerr) register(pm, 'Balanced Error', calcbal) def accuracy(out, lab): """ Computes Accuracy. Expects labels to be +/-1 and predictions sign(output) """ return numpy.mean(numpy.sign(out) == lab) register(pm, 'Accuracy', accuracy) def error(out, lab): """ Computes Error Rate. Expects labels to be +/-1 and predictions sign(output) """ return numpy.mean(numpy.sign(out) != lab) register(pm, 'Error', error) def get_tp(out, lab): """ Computes True Positive Rate. Expects labels to be +/-1 and predictions sign(output) """ pidx=(lab==+1) return numpy.sum(numpy.sign(out[pidx])>0) register(pm,'True Positive Rate', get_tp) def get_tn(out, lab): """ Computes True Negative Rate. Expects labels to be +/-1 and predictions sign(output) """ nidx=(lab==-1) return numpy.sum(numpy.sign(out[nidx])<0) register(pm, 'True Negative Rate', get_tn) def get_fp(out, lab): """ Computes False Positive Rate. Expects labels to be +/-1 and predictions sign(output) """ nidx=(lab==-1) return numpy.sum(numpy.sign(out[nidx])>0) register(pm, 'False Positive Rate', get_fp) def get_fn(out, lab): """ Computes False Negative Rate. Expects labels to be +/-1 and predictions sign(output) """ pidx=(lab==+1) return numpy.sum(numpy.sign(out[pidx])<0) register(pm, 'False Negative Rate', get_fn) def calcwracc(out, lab): """ Computes Weighted Relative Accuracy. Expects labels to be +/-1 and predictions sign(output) """ tp=get_tp(out, lab) fp=get_fp(out, lab) tn=get_tn(out, lab) fn=get_fn(out, lab) return tp/(tp+fn) - fp/(fp+tn) register(pm, 'Weighted Relative Accuracy', calcwracc) def calcf1(out, lab): """ Computes the F1 score. Expects labels to be +/-1 and predictions sign(output) """ tp=get_tp(out, lab) fp=get_fp(out, lab) fn=get_fn(out, lab) return 2.0*tp/(2.0*tp+fp+fn) register(pm, 'F1 Score', calcf1) def calccc(out, lab): """ Computes the cross correlation coefficient. Expects labels to be +/-1 and predictions sign(output) """ tp=float(get_tp(out, lab)) fp=float(get_fp(out, lab)) tn=float(get_tn(out, lab)) fn=float(get_fn(out, lab)) return (tp*tn-fp*fn)/numpy.sqrt((tp+fp)*(tp+fn)*(tn+fp)*(tn+fn)) register(pm, 'Cross Correlation Coefficient', calccc) def calcroc(out, lab): """ Computes the ROC curve. Expects labels to be +/-1 and real-valued predictions """ idx = numpy.argsort(out) tp=numpy.concatenate( ([1], 1-numpy.cumsum(lab[idx]>0)/float(numpy.sum(lab > 0))) ) fp=numpy.concatenate( ([1], 1-numpy.cumsum(lab[idx]<0)/float(numpy.sum(lab < 0))) ) score = 0.5*numpy.sum((fp[:-1]-fp[1:])*(tp[:-1]+tp[1:])) curve= {'x' : fp, 'y' : tp, 'x_name' : 'False Positive Rate', 'y_name' : 'True Positive Rate'} return (score, 'Curve', curve) register(pm, 'ROC Curve', calcroc) def calcrocscore(out, lab): """ Computes the area under the ROC curve. Expects labels to be +/-1 and real-valued predictions """ return calcroc(out,lab)[0] register(pm, 'Area under ROC Curve', calcrocscore) def calcprc(out, lab): """ Computes the precision recall curve. Expects labels to be +/-1 and real-valued predictions """ pmap=(lab==1) idx = numpy.argsort(-out) pmap=pmap[idx] recall=numpy.cumsum(pmap)/numpy.double(sum(lab==1)) precision=numpy.cumsum(pmap)/(numpy.arange(len(pmap),dtype=numpy.double)+1.0) score = 0.5*numpy.sum((precision[:-1]+precision[1:])*(recall[1:]-recall[:-1])) curve= {'x' : precision, 'y' : recall, 'x_name' : 'Precision', 'y_name' : 'Recall'} return (score, 'Curve', curve) register(pm, 'Precision Recall Curve', calcprc) def calcprcscore(out, lab): """ Computes the area under the precision recall curve. Expects labels to be +/-1 and real-valued predictions """ return calcprc(out,lab)[0] register(pm, 'Area under Precision Recall Curve', calcprcscore) mldata-utils-0.4.1/mleval/__init__.py0000600000175000017500000000024311474776561016544 0ustar sonnesonne""" Utility functions to perform evaluations, mostly for mldata.org. """ __all__ = ['evaluation', 'parse', 'multiclass', 'classification', 'regression','output'] mldata-utils-0.4.1/mleval/output.py0000600000175000017500000000601511474776561016350 0ustar sonnesonne"""A class to deal with standard prediction outputs. Currently supports: - binary classification - multiclass - regression """ from numpy import array, sign, unique class Outputs(object): """A container for the labels/predictions of a supervised learning method """ def str2label(str_label): raise NotImplementedError def label2str(preds): raise NotImplementedError class OutputsClassification(Outputs): """A container for labels/predictions for classification""" name2id = {} id2name = {} def str2label(self, str_label): """Convert labels into class ids""" self.init_id(str_label) int_label = [] for lab in str_label: int_label.append(self.name2id[str(lab[0])]) return array(int_label, dtype=float) class OutputsBinClass(OutputsClassification): """A container for labels/predictions for binary classification """ def init_id(self, str_label): """Discover the names of the labels""" all_names = unique(str_label) all_names = map(str, all_names) assert(len(all_names) == 2) self.name2id = {all_names[0]: 1, all_names[1]: -1, } self.id2name = {1: all_names[0], -1: all_names[1], } def label2str(self, preds): """Convert predictions into classification labels""" pred_label = [] for pred in preds: pred_label.append(self.id2name[sign(pred)]) return pred_label class OutputsMultiClass(OutputsClassification): """A container for labels/predictions for multiclass classification """ def init_id(self, str_label): """Discover the names of the labels""" all_names = unique(str_label) all_names = map(str, all_names) assert(len(all_names) > 2) self.name2id = {} self.id2name = {} for (idx, name) in enumerate(all_names): self.name2id[name] = idx self.id2name[idx] = name def label2str(self, preds): """Convert predictions into classification labels""" pred_label = [] for pred in preds: pred_label.append(self.id2name[pred]) return pred_label class OutputsRegression(Outputs): """A container for labels/predictions for regression """ def str2label(self, str_label): """Convert labels into floats""" float_label = [] for lab in str_label: float_label.append(float(lab[0])) return array(float_label, dtype=float) def label2str(self, preds): """Convert predictions into strings""" return map(str, preds) def init_output(task_type): """A factory for creating the right output class""" output_types = {'Binary Classification': OutputsBinClass, 'Multi Class Classification': OutputsMultiClass, 'Regression': OutputsRegression, } return output_types[task_type]() mldata-utils-0.4.1/mleval/parse.py0000600000175000017500000000421711450606061016101 0ustar sonnesonneimport numpy def parse_floats(fil): """ Parse file and return numpy array """ fil.seek(0) values=list() for l in fil.xreadlines(): values.append( [ float(e) for e in l.split() ] ) return numpy.array(values) def parse_strings(output_file, label_file): """ Parse file expecting a few string objects only converting them to +/-1 and return numpy array """ output_file.seek(0) label_file.seek(0) output_valid_values=set() label_valid_values=set() for l in output_file.xreadlines(): output_valid_values.add(l.strip()) for l in label_file.xreadlines(): label_valid_values.add(l.strip()) if len(output_valid_values)>2 or len(label_valid_values)>2: msg="Only Binary outputs are currently supported but found more than two:\n" msg+="outputs are: %s\n" % str(output_valid_values) msg+="labels are: %s\n" % str(label_valid_values) raise Exception(msg) if len(output_valid_values-label_valid_values): msg="Invalid outputs found\n" msg+="outputs are: %s\n" % str(output_valid_values) msg+="labels are: %s\n" % str(label_valid_values) raise Exception(msg) if len(output_valid_values-label_valid_values): msg="Invalid outputs found\n" msg+="outputs are: %s\n" % str(output_valid_values) msg+="labels are: %s\n" % str(label_valid_values) raise Exception(msg) map_dict=dict() i=-1 for k in label_valid_values: map_dict[k]=i i+=1 output_file.seek(0) label_file.seek(0) output_values=list() for l in output_file.xreadlines(): output_values.append( [ float(map_dict[e.strip()]) for e in l.split() ] ) label_values=list() for l in label_file.xreadlines(): label_values.append( [ float(map_dict[e.strip()]) for e in l.split() ] ) return numpy.array(output_values),numpy.array(label_values) def brute_force_parse(output_file, label_file): try: outputs=parse_floats(output_file) labels=parse_floats(label_file) return outputs,labels except: return parse_strings(output_file, label_file) mldata-utils-0.4.1/mleval/regression.py0000600000175000017500000000116711451413765017160 0ustar sonnesonneimport numpy from util import register pm = dict() # regression measures def mean_absolute_error(out, truth): """ Regression evaluation using Mean Absolute Error Expects out and truth to be real valued vectors """ return numpy.mean(numpy.abs(out-truth)) register(pm, 'Mean Absolute Error', mean_absolute_error) def root_mean_squared_error(out, truth): """ Regression evaluation using Root Mean Squared Error (RMSE) Expects out and truth to be real valued vectors """ return numpy.sqrt(numpy.mean((out-truth)**2)) register(pm, 'Root Mean Squared Error', root_mean_squared_error) mldata-utils-0.4.1/mleval/evaluation.py0000644000175000017500000000127011451415525017146 0ustar sonnesonneimport multiclass import classification import regression """ dictionaries of known performance measures: keys are human readable names, values are tuples of (functioname, description) pm_hierarchy - is a dictionary of tasks specific perf measures (e.g. for Classification, Regression, ...). To add a new measure just add a function, e.g. def calcbal(out, lab): "Computes the Balanced Error" ... and add it to the the appropriate dictionary register(pm, 'Balanced Error', calcbal) """ pm_hierarchy = { 'Regression': regression.pm, 'Binary Classification': classification.pm, 'Multi Class Classification' : multiclass.pm} mldata-utils-0.4.1/mleval/multiclass.py0000600000175000017500000000064311452347246017157 0ustar sonnesonneimport numpy from util import register pm = dict() def accuracy(out, lab): """ Computes Accuracy. Expects labels and predictions to be integers """ return numpy.mean(out == lab) register(pm, 'Accuracy', accuracy) def error(out, lab): """ Computes Error Rate. Expects labels and predictions to be integers """ return numpy.mean(out != lab) register(pm, 'Error', error) mldata-utils-0.4.1/scripts/0000700000175000017500000000000011477311160014620 5ustar sonnesonnemldata-utils-0.4.1/scripts/ml2h50000700000175000017500000000721011473010123015464 0ustar sonnesonne#!/usr/bin/env python """ Convert from and to HDF5 """ import sys, os, getopt import ml2h5.converter from ml2h5.converter import TO_H5,FROM_H5 from optparse import OptionParser def oParse(): parser=OptionParser() parser.add_option('-s','--seperator',dest='seperator',help='Seperator to use to seperate variables in examples. Default is ","',default=None) parser.add_option('-v', '--verify',action='store_true',dest='verify',help='Verify the converted data',default=False) parser.add_option('-f', '--first-line-attribute-names',action='store_true',dest='attribute_names_first',help='First line contains attributes names (for CSV)',default=False) parser.add_option('-t', '--type',dest='type',help='File type to convert, can either be data or task',default='data') parser.add_option('-i', '--format-in',dest='format_in',help='File format of in-file (if auto-detection fails)',default=None) parser.add_option('-o', '--format-out',dest='format_out',help='File format of out-file (if auto-detection fails)',default=None) parser.add_option('-m', '--merge-data',action='store_true',dest='merge',help='',default=False) parser.add_option('-n', '--no-conversion',action='store_false',dest='convert',help='',default=True) msg=[sys.argv[0] + """ [options] ") Convert to . The following types are supported:"""] for item in TO_H5: msg.append(' ' + item + ' -> ' + 'h5') msg.append('') for item in FROM_H5: msg.append(' h5 -> ' + item) parser.set_usage("\n".join(msg)) return parser class Options: """Option. Should not be instantiated. @cvar seperator: seperator to seperate variables in examples @type output: string @cvar verify: if converted data shall be verified against input data @type verify: boolean @cvar attribute_names_first: if first line in CSV files shall be treated as attribute names @type attribute_names_first: boolean @cvar format_in: file format of in-file @type format_in: string @cvar format_out: file format of out-file @type format_out: string """ seperator = None verify = False attribute_names_first = False format_in = None format_out = None merge = False type = None def rm_opt(option, value=None): """Remove given option and value from sys.argv. @param option: option to remove @type option: string @param value: value to remove @type value: string """ if not value: sys.argv.remove(option) return try: sys.argv.remove(option + value) except ValueError: try: sys.argv.remove(option) sys.argv.remove(value) except ValueError: sys.argv.remove(option + '=' + value) if __name__ == "__main__": op=oParse() (Options,args)=op.parse_args() if len(args)!=2: op.print_help() sys.exit(1) if Options.format_in == 'h5': seperator = None elif Options.seperator: seperator = Options.seperator else: seperator = ml2h5.fileformat.infer_seperator(args[0]) try: c = ml2h5.converter.Converter( args[0], args[1], format_in=Options.format_in, format_out=Options.format_out, seperator=seperator, attribute_names_first=Options.attribute_names_first, merge=Options.merge, type=Options.type ) if Options.convert: c.run(verify=Options.verify) if Options.verify: c.verify() except ml2h5.converter.ConversionError, e: e.print_error() sys.exit(1) mldata-utils-0.4.1/scripts/mleval0000700000175000017500000000414511451047767016045 0ustar sonnesonne#!/usr/bin/env python """ Perform evaluation """ import sys, os, getopt import mleval from mleval.evaluation import pm_hierarchy, pm from mleval.parse import brute_force_parse def usage(): msg = [mleval.PROGRAM_NAME + ', Version ' + mleval.VERSION + """ Usage: """ + sys.argv[0] + """ [options] Options: --pm="" Performance Measure Supported performance measures are: """] for pm_type in pm_hierarchy.keys(): p=pm_hierarchy[pm_type] msg.append(' %s\n %s\n\n' % (pm_type, len(pm_type)*'=')) for k in p.keys(): msg.append(' "%s"\n' % k) msg.append(' %s\n\n' % p[k][1]) print ''.join(msg) def parse_options(): """Parse given options.""" try: opts, args = getopt.getopt(sys.argv[1:], '', ['pm=']) except getopt.GetoptError, err: # print help information and exit print str(err) + "\n" usage() sys.exit(1) perf_measure=None for o, a in opts: if o in ('--pm'): perf_measure = a else: print 'Unhandled option: ', o sys.exit(2) if len(args)!=2 or not perf_measure: usage() sys.exit(1) # pm, outputs, true labels return (perf_measure, args[0], args[1]) if __name__ == "__main__": argc = len(sys.argv) if argc < 3: usage() sys.exit(1) perf_measure, outputfile, labelfile = parse_options() try: of=file(outputfile) except IOError: print "could not open output_file '%s'" % outputfile sys.exit(1) try: lf=file(labelfile) except IOError: print "could not open truth_file '%s'" % labelfile sys.exit(1) if not pm.has_key(perf_measure): print "not a valid performance measure" sys.exit(1) try: outputs, labels=brute_force_parse(of, lf) except Exception, err: print err sys.exit(2) if len(outputs)!=len(labels): print "outputs and label length don't match" sys.exit(2) print pm[perf_measure][0](outputs, labels) mldata-utils-0.4.1/scripts/ml2h5extract_data0000700000175000017500000000053311450541742020064 0ustar sonnesonne#!/usr/bin/env python """ Print an extract of an HDF5 file, similar to website output """ import sys import ml2h5.data def usage(): print """Usage: """ + sys.argv[0] + """ """ if __name__ == "__main__": argc = len(sys.argv) if argc < 2: usage() sys.exit(1) print ml2h5.data.get_extract(sys.argv[1]) mldata-utils-0.4.1/scripts/ml2h5extract_task0000700000175000017500000000053311450541742020115 0ustar sonnesonne#!/usr/bin/env python """ Print an extract of an HDF5 file, similar to website output """ import sys import ml2h5.task def usage(): print """Usage: """ + sys.argv[0] + """ """ if __name__ == "__main__": argc = len(sys.argv) if argc < 2: usage() sys.exit(1) print ml2h5.task.get_extract(sys.argv[1]) mldata-utils-0.4.1/setup.py0000600000175000017500000000405111472426767014664 0ustar sonnesonnefrom distutils.core import setup VERSION='0.4.1' DESCRIPTION='Utilities around hdf5 and for compute performance measures', LONG_DESCRIPTION=""" Utility functions to perform evaluations and to convert from/to the hdf5 based mldata format into/from typical machine learning formats like .csv,.arff,.octave,.mat. This code is actually used on mldata.org to perform the evaluations and data and task conversions. """ AUTHOR = ['Soeren Sonnenburg', 'Sebastian Henschel', 'Mikio Braun', 'Hagen Zahn','Cheng Soon Ong'] AUTHOR_EMAIL = ['Soeren.Sonnenburg@tu-berlin.de', 'mldata@kodeaffe.de', 'mikio@cs.tu-berlin.de', 'hzahn@informatik.hu-berlin.de', 'chengsoon.ong@inf.ethz.ch'] URL = 'http://mloss.org/software/view/262/' LICENSE = """ This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; version 3 of the License. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, check http://www.gnu.org/licenses/gpl-3.0.html """ setup( name='mldata-utils', version=VERSION, description=DESCRIPTION, long_description=LONG_DESCRIPTION, maintainer=AUTHOR, maintainer_email=AUTHOR_EMAIL, license=LICENSE, url=URL, packages=['ml2h5', 'ml2h5.converter', 'mleval'], scripts=['scripts/ml2h5', 'scripts/ml2h5extract_data', 'scripts/ml2h5extract_task', 'scripts/mleval'], classifiers=[ 'Development Status :: 3 - Alpha', 'Environment :: Console', 'Intended Audience :: Science/Research', 'License :: OSI Approved :: GNU General Public License (GPL)', 'Natural Language :: English', 'Operating System :: POSIX', 'Programming Language :: Python', 'Topic :: Utilities', ], platforms=['POSIX'], ) mldata-utils-0.4.1/PKG-INFO0000600000175000017500000000322511477311160014232 0ustar sonnesonneMetadata-Version: 1.0 Name: mldata-utils Version: 0.4.1 Summary: ('Utilities around hdf5 and for compute performance measures',) Home-page: http://mloss.org/software/view/262/ Author: ['Soeren Sonnenburg', 'Sebastian Henschel', 'Mikio Braun', 'Hagen Zahn', 'Cheng Soon Ong'] Author-email: ['Soeren.Sonnenburg@tu-berlin.de', 'mldata@kodeaffe.de', 'mikio@cs.tu-berlin.de', 'hzahn@informatik.hu-berlin.de', 'chengsoon.ong@inf.ethz.ch'] License: This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; version 3 of the License. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, check http://www.gnu.org/licenses/gpl-3.0.html Description: Utility functions to perform evaluations and to convert from/to the hdf5 based mldata format into/from typical machine learning formats like .csv,.arff,.octave,.mat. This code is actually used on mldata.org to perform the evaluations and data and task conversions. Platform: POSIX Classifier: Development Status :: 3 - Alpha Classifier: Environment :: Console Classifier: Intended Audience :: Science/Research Classifier: License :: OSI Approved :: GNU General Public License (GPL) Classifier: Natural Language :: English Classifier: Operating System :: POSIX Classifier: Programming Language :: Python Classifier: Topic :: Utilities mldata-utils-0.4.1/COPYING0000600000175000017500000010451311450541742014174 0ustar sonnesonne GNU GENERAL PUBLIC LICENSE Version 3, 29 June 2007 Copyright (C) 2007 Free Software Foundation, Inc. Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. Preamble The GNU General Public License is a free, copyleft license for software and other kinds of works. The licenses for most software and other practical works are designed to take away your freedom to share and change the works. By contrast, the GNU General Public License is intended to guarantee your freedom to share and change all versions of a program--to make sure it remains free software for all its users. We, the Free Software Foundation, use the GNU General Public License for most of our software; it applies also to any other work released this way by its authors. You can apply it to your programs, too. When we speak of free software, we are referring to freedom, not price. Our General Public Licenses are designed to make sure that you have the freedom to distribute copies of free software (and charge for them if you wish), that you receive source code or can get it if you want it, that you can change the software or use pieces of it in new free programs, and that you know you can do these things. To protect your rights, we need to prevent others from denying you these rights or asking you to surrender the rights. Therefore, you have certain responsibilities if you distribute copies of the software, or if you modify it: responsibilities to respect the freedom of others. For example, if you distribute copies of such a program, whether gratis or for a fee, you must pass on to the recipients the same freedoms that you received. You must make sure that they, too, receive or can get the source code. And you must show them these terms so they know their rights. Developers that use the GNU GPL protect your rights with two steps: (1) assert copyright on the software, and (2) offer you this License giving you legal permission to copy, distribute and/or modify it. For the developers' and authors' protection, the GPL clearly explains that there is no warranty for this free software. For both users' and authors' sake, the GPL requires that modified versions be marked as changed, so that their problems will not be attributed erroneously to authors of previous versions. Some devices are designed to deny users access to install or run modified versions of the software inside them, although the manufacturer can do so. This is fundamentally incompatible with the aim of protecting users' freedom to change the software. The systematic pattern of such abuse occurs in the area of products for individuals to use, which is precisely where it is most unacceptable. Therefore, we have designed this version of the GPL to prohibit the practice for those products. If such problems arise substantially in other domains, we stand ready to extend this provision to those domains in future versions of the GPL, as needed to protect the freedom of users. Finally, every program is threatened constantly by software patents. States should not allow patents to restrict development and use of software on general-purpose computers, but in those that do, we wish to avoid the special danger that patents applied to a free program could make it effectively proprietary. To prevent this, the GPL assures that patents cannot be used to render the program non-free. The precise terms and conditions for copying, distribution and modification follow. TERMS AND CONDITIONS 0. Definitions. "This License" refers to version 3 of the GNU General Public License. "Copyright" also means copyright-like laws that apply to other kinds of works, such as semiconductor masks. "The Program" refers to any copyrightable work licensed under this License. Each licensee is addressed as "you". "Licensees" and "recipients" may be individuals or organizations. To "modify" a work means to copy from or adapt all or part of the work in a fashion requiring copyright permission, other than the making of an exact copy. The resulting work is called a "modified version" of the earlier work or a work "based on" the earlier work. A "covered work" means either the unmodified Program or a work based on the Program. To "propagate" a work means to do anything with it that, without permission, would make you directly or secondarily liable for infringement under applicable copyright law, except executing it on a computer or modifying a private copy. Propagation includes copying, distribution (with or without modification), making available to the public, and in some countries other activities as well. To "convey" a work means any kind of propagation that enables other parties to make or receive copies. Mere interaction with a user through a computer network, with no transfer of a copy, is not conveying. An interactive user interface displays "Appropriate Legal Notices" to the extent that it includes a convenient and prominently visible feature that (1) displays an appropriate copyright notice, and (2) tells the user that there is no warranty for the work (except to the extent that warranties are provided), that licensees may convey the work under this License, and how to view a copy of this License. If the interface presents a list of user commands or options, such as a menu, a prominent item in the list meets this criterion. 1. Source Code. The "source code" for a work means the preferred form of the work for making modifications to it. "Object code" means any non-source form of a work. A "Standard Interface" means an interface that either is an official standard defined by a recognized standards body, or, in the case of interfaces specified for a particular programming language, one that is widely used among developers working in that language. The "System Libraries" of an executable work include anything, other than the work as a whole, that (a) is included in the normal form of packaging a Major Component, but which is not part of that Major Component, and (b) serves only to enable use of the work with that Major Component, or to implement a Standard Interface for which an implementation is available to the public in source code form. A "Major Component", in this context, means a major essential component (kernel, window system, and so on) of the specific operating system (if any) on which the executable work runs, or a compiler used to produce the work, or an object code interpreter used to run it. The "Corresponding Source" for a work in object code form means all the source code needed to generate, install, and (for an executable work) run the object code and to modify the work, including scripts to control those activities. However, it does not include the work's System Libraries, or general-purpose tools or generally available free programs which are used unmodified in performing those activities but which are not part of the work. For example, Corresponding Source includes interface definition files associated with source files for the work, and the source code for shared libraries and dynamically linked subprograms that the work is specifically designed to require, such as by intimate data communication or control flow between those subprograms and other parts of the work. The Corresponding Source need not include anything that users can regenerate automatically from other parts of the Corresponding Source. The Corresponding Source for a work in source code form is that same work. 2. Basic Permissions. All rights granted under this License are granted for the term of copyright on the Program, and are irrevocable provided the stated conditions are met. This License explicitly affirms your unlimited permission to run the unmodified Program. The output from running a covered work is covered by this License only if the output, given its content, constitutes a covered work. This License acknowledges your rights of fair use or other equivalent, as provided by copyright law. You may make, run and propagate covered works that you do not convey, without conditions so long as your license otherwise remains in force. You may convey covered works to others for the sole purpose of having them make modifications exclusively for you, or provide you with facilities for running those works, provided that you comply with the terms of this License in conveying all material for which you do not control copyright. Those thus making or running the covered works for you must do so exclusively on your behalf, under your direction and control, on terms that prohibit them from making any copies of your copyrighted material outside their relationship with you. Conveying under any other circumstances is permitted solely under the conditions stated below. Sublicensing is not allowed; section 10 makes it unnecessary. 3. Protecting Users' Legal Rights From Anti-Circumvention Law. No covered work shall be deemed part of an effective technological measure under any applicable law fulfilling obligations under article 11 of the WIPO copyright treaty adopted on 20 December 1996, or similar laws prohibiting or restricting circumvention of such measures. When you convey a covered work, you waive any legal power to forbid circumvention of technological measures to the extent such circumvention is effected by exercising rights under this License with respect to the covered work, and you disclaim any intention to limit operation or modification of the work as a means of enforcing, against the work's users, your or third parties' legal rights to forbid circumvention of technological measures. 4. Conveying Verbatim Copies. You may convey verbatim copies of the Program's source code as you receive it, in any medium, provided that you conspicuously and appropriately publish on each copy an appropriate copyright notice; keep intact all notices stating that this License and any non-permissive terms added in accord with section 7 apply to the code; keep intact all notices of the absence of any warranty; and give all recipients a copy of this License along with the Program. You may charge any price or no price for each copy that you convey, and you may offer support or warranty protection for a fee. 5. Conveying Modified Source Versions. You may convey a work based on the Program, or the modifications to produce it from the Program, in the form of source code under the terms of section 4, provided that you also meet all of these conditions: a) The work must carry prominent notices stating that you modified it, and giving a relevant date. b) The work must carry prominent notices stating that it is released under this License and any conditions added under section 7. This requirement modifies the requirement in section 4 to "keep intact all notices". c) You must license the entire work, as a whole, under this License to anyone who comes into possession of a copy. This License will therefore apply, along with any applicable section 7 additional terms, to the whole of the work, and all its parts, regardless of how they are packaged. This License gives no permission to license the work in any other way, but it does not invalidate such permission if you have separately received it. d) If the work has interactive user interfaces, each must display Appropriate Legal Notices; however, if the Program has interactive interfaces that do not display Appropriate Legal Notices, your work need not make them do so. A compilation of a covered work with other separate and independent works, which are not by their nature extensions of the covered work, and which are not combined with it such as to form a larger program, in or on a volume of a storage or distribution medium, is called an "aggregate" if the compilation and its resulting copyright are not used to limit the access or legal rights of the compilation's users beyond what the individual works permit. Inclusion of a covered work in an aggregate does not cause this License to apply to the other parts of the aggregate. 6. Conveying Non-Source Forms. You may convey a covered work in object code form under the terms of sections 4 and 5, provided that you also convey the machine-readable Corresponding Source under the terms of this License, in one of these ways: a) Convey the object code in, or embodied in, a physical product (including a physical distribution medium), accompanied by the Corresponding Source fixed on a durable physical medium customarily used for software interchange. b) Convey the object code in, or embodied in, a physical product (including a physical distribution medium), accompanied by a written offer, valid for at least three years and valid for as long as you offer spare parts or customer support for that product model, to give anyone who possesses the object code either (1) a copy of the Corresponding Source for all the software in the product that is covered by this License, on a durable physical medium customarily used for software interchange, for a price no more than your reasonable cost of physically performing this conveying of source, or (2) access to copy the Corresponding Source from a network server at no charge. c) Convey individual copies of the object code with a copy of the written offer to provide the Corresponding Source. This alternative is allowed only occasionally and noncommercially, and only if you received the object code with such an offer, in accord with subsection 6b. d) Convey the object code by offering access from a designated place (gratis or for a charge), and offer equivalent access to the Corresponding Source in the same way through the same place at no further charge. You need not require recipients to copy the Corresponding Source along with the object code. If the place to copy the object code is a network server, the Corresponding Source may be on a different server (operated by you or a third party) that supports equivalent copying facilities, provided you maintain clear directions next to the object code saying where to find the Corresponding Source. Regardless of what server hosts the Corresponding Source, you remain obligated to ensure that it is available for as long as needed to satisfy these requirements. e) Convey the object code using peer-to-peer transmission, provided you inform other peers where the object code and Corresponding Source of the work are being offered to the general public at no charge under subsection 6d. A separable portion of the object code, whose source code is excluded from the Corresponding Source as a System Library, need not be included in conveying the object code work. A "User Product" is either (1) a "consumer product", which means any tangible personal property which is normally used for personal, family, or household purposes, or (2) anything designed or sold for incorporation into a dwelling. In determining whether a product is a consumer product, doubtful cases shall be resolved in favor of coverage. For a particular product received by a particular user, "normally used" refers to a typical or common use of that class of product, regardless of the status of the particular user or of the way in which the particular user actually uses, or expects or is expected to use, the product. A product is a consumer product regardless of whether the product has substantial commercial, industrial or non-consumer uses, unless such uses represent the only significant mode of use of the product. "Installation Information" for a User Product means any methods, procedures, authorization keys, or other information required to install and execute modified versions of a covered work in that User Product from a modified version of its Corresponding Source. The information must suffice to ensure that the continued functioning of the modified object code is in no case prevented or interfered with solely because modification has been made. If you convey an object code work under this section in, or with, or specifically for use in, a User Product, and the conveying occurs as part of a transaction in which the right of possession and use of the User Product is transferred to the recipient in perpetuity or for a fixed term (regardless of how the transaction is characterized), the Corresponding Source conveyed under this section must be accompanied by the Installation Information. But this requirement does not apply if neither you nor any third party retains the ability to install modified object code on the User Product (for example, the work has been installed in ROM). The requirement to provide Installation Information does not include a requirement to continue to provide support service, warranty, or updates for a work that has been modified or installed by the recipient, or for the User Product in which it has been modified or installed. Access to a network may be denied when the modification itself materially and adversely affects the operation of the network or violates the rules and protocols for communication across the network. Corresponding Source conveyed, and Installation Information provided, in accord with this section must be in a format that is publicly documented (and with an implementation available to the public in source code form), and must require no special password or key for unpacking, reading or copying. 7. Additional Terms. "Additional permissions" are terms that supplement the terms of this License by making exceptions from one or more of its conditions. Additional permissions that are applicable to the entire Program shall be treated as though they were included in this License, to the extent that they are valid under applicable law. If additional permissions apply only to part of the Program, that part may be used separately under those permissions, but the entire Program remains governed by this License without regard to the additional permissions. When you convey a copy of a covered work, you may at your option remove any additional permissions from that copy, or from any part of it. (Additional permissions may be written to require their own removal in certain cases when you modify the work.) You may place additional permissions on material, added by you to a covered work, for which you have or can give appropriate copyright permission. Notwithstanding any other provision of this License, for material you add to a covered work, you may (if authorized by the copyright holders of that material) supplement the terms of this License with terms: a) Disclaiming warranty or limiting liability differently from the terms of sections 15 and 16 of this License; or b) Requiring preservation of specified reasonable legal notices or author attributions in that material or in the Appropriate Legal Notices displayed by works containing it; or c) Prohibiting misrepresentation of the origin of that material, or requiring that modified versions of such material be marked in reasonable ways as different from the original version; or d) Limiting the use for publicity purposes of names of licensors or authors of the material; or e) Declining to grant rights under trademark law for use of some trade names, trademarks, or service marks; or f) Requiring indemnification of licensors and authors of that material by anyone who conveys the material (or modified versions of it) with contractual assumptions of liability to the recipient, for any liability that these contractual assumptions directly impose on those licensors and authors. All other non-permissive additional terms are considered "further restrictions" within the meaning of section 10. If the Program as you received it, or any part of it, contains a notice stating that it is governed by this License along with a term that is a further restriction, you may remove that term. If a license document contains a further restriction but permits relicensing or conveying under this License, you may add to a covered work material governed by the terms of that license document, provided that the further restriction does not survive such relicensing or conveying. If you add terms to a covered work in accord with this section, you must place, in the relevant source files, a statement of the additional terms that apply to those files, or a notice indicating where to find the applicable terms. Additional terms, permissive or non-permissive, may be stated in the form of a separately written license, or stated as exceptions; the above requirements apply either way. 8. Termination. You may not propagate or modify a covered work except as expressly provided under this License. Any attempt otherwise to propagate or modify it is void, and will automatically terminate your rights under this License (including any patent licenses granted under the third paragraph of section 11). However, if you cease all violation of this License, then your license from a particular copyright holder is reinstated (a) provisionally, unless and until the copyright holder explicitly and finally terminates your license, and (b) permanently, if the copyright holder fails to notify you of the violation by some reasonable means prior to 60 days after the cessation. Moreover, your license from a particular copyright holder is reinstated permanently if the copyright holder notifies you of the violation by some reasonable means, this is the first time you have received notice of violation of this License (for any work) from that copyright holder, and you cure the violation prior to 30 days after your receipt of the notice. Termination of your rights under this section does not terminate the licenses of parties who have received copies or rights from you under this License. If your rights have been terminated and not permanently reinstated, you do not qualify to receive new licenses for the same material under section 10. 9. Acceptance Not Required for Having Copies. You are not required to accept this License in order to receive or run a copy of the Program. Ancillary propagation of a covered work occurring solely as a consequence of using peer-to-peer transmission to receive a copy likewise does not require acceptance. However, nothing other than this License grants you permission to propagate or modify any covered work. These actions infringe copyright if you do not accept this License. Therefore, by modifying or propagating a covered work, you indicate your acceptance of this License to do so. 10. Automatic Licensing of Downstream Recipients. Each time you convey a covered work, the recipient automatically receives a license from the original licensors, to run, modify and propagate that work, subject to this License. You are not responsible for enforcing compliance by third parties with this License. An "entity transaction" is a transaction transferring control of an organization, or substantially all assets of one, or subdividing an organization, or merging organizations. If propagation of a covered work results from an entity transaction, each party to that transaction who receives a copy of the work also receives whatever licenses to the work the party's predecessor in interest had or could give under the previous paragraph, plus a right to possession of the Corresponding Source of the work from the predecessor in interest, if the predecessor has it or can get it with reasonable efforts. You may not impose any further restrictions on the exercise of the rights granted or affirmed under this License. For example, you may not impose a license fee, royalty, or other charge for exercise of rights granted under this License, and you may not initiate litigation (including a cross-claim or counterclaim in a lawsuit) alleging that any patent claim is infringed by making, using, selling, offering for sale, or importing the Program or any portion of it. 11. Patents. A "contributor" is a copyright holder who authorizes use under this License of the Program or a work on which the Program is based. The work thus licensed is called the contributor's "contributor version". A contributor's "essential patent claims" are all patent claims owned or controlled by the contributor, whether already acquired or hereafter acquired, that would be infringed by some manner, permitted by this License, of making, using, or selling its contributor version, but do not include claims that would be infringed only as a consequence of further modification of the contributor version. For purposes of this definition, "control" includes the right to grant patent sublicenses in a manner consistent with the requirements of this License. Each contributor grants you a non-exclusive, worldwide, royalty-free patent license under the contributor's essential patent claims, to make, use, sell, offer for sale, import and otherwise run, modify and propagate the contents of its contributor version. In the following three paragraphs, a "patent license" is any express agreement or commitment, however denominated, not to enforce a patent (such as an express permission to practice a patent or covenant not to sue for patent infringement). To "grant" such a patent license to a party means to make such an agreement or commitment not to enforce a patent against the party. If you convey a covered work, knowingly relying on a patent license, and the Corresponding Source of the work is not available for anyone to copy, free of charge and under the terms of this License, through a publicly available network server or other readily accessible means, then you must either (1) cause the Corresponding Source to be so available, or (2) arrange to deprive yourself of the benefit of the patent license for this particular work, or (3) arrange, in a manner consistent with the requirements of this License, to extend the patent license to downstream recipients. "Knowingly relying" means you have actual knowledge that, but for the patent license, your conveying the covered work in a country, or your recipient's use of the covered work in a country, would infringe one or more identifiable patents in that country that you have reason to believe are valid. If, pursuant to or in connection with a single transaction or arrangement, you convey, or propagate by procuring conveyance of, a covered work, and grant a patent license to some of the parties receiving the covered work authorizing them to use, propagate, modify or convey a specific copy of the covered work, then the patent license you grant is automatically extended to all recipients of the covered work and works based on it. A patent license is "discriminatory" if it does not include within the scope of its coverage, prohibits the exercise of, or is conditioned on the non-exercise of one or more of the rights that are specifically granted under this License. You may not convey a covered work if you are a party to an arrangement with a third party that is in the business of distributing software, under which you make payment to the third party based on the extent of your activity of conveying the work, and under which the third party grants, to any of the parties who would receive the covered work from you, a discriminatory patent license (a) in connection with copies of the covered work conveyed by you (or copies made from those copies), or (b) primarily for and in connection with specific products or compilations that contain the covered work, unless you entered into that arrangement, or that patent license was granted, prior to 28 March 2007. Nothing in this License shall be construed as excluding or limiting any implied license or other defenses to infringement that may otherwise be available to you under applicable patent law. 12. No Surrender of Others' Freedom. If conditions are imposed on you (whether by court order, agreement or otherwise) that contradict the conditions of this License, they do not excuse you from the conditions of this License. If you cannot convey a covered work so as to satisfy simultaneously your obligations under this License and any other pertinent obligations, then as a consequence you may not convey it at all. For example, if you agree to terms that obligate you to collect a royalty for further conveying from those to whom you convey the Program, the only way you could satisfy both those terms and this License would be to refrain entirely from conveying the Program. 13. Use with the GNU Affero General Public License. Notwithstanding any other provision of this License, you have permission to link or combine any covered work with a work licensed under version 3 of the GNU Affero General Public License into a single combined work, and to convey the resulting work. The terms of this License will continue to apply to the part which is the covered work, but the special requirements of the GNU Affero General Public License, section 13, concerning interaction through a network will apply to the combination as such. 14. Revised Versions of this License. The Free Software Foundation may publish revised and/or new versions of the GNU General Public License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. Each version is given a distinguishing version number. If the Program specifies that a certain numbered version of the GNU General Public License "or any later version" applies to it, you have the option of following the terms and conditions either of that numbered version or of any later version published by the Free Software Foundation. If the Program does not specify a version number of the GNU General Public License, you may choose any version ever published by the Free Software Foundation. If the Program specifies that a proxy can decide which future versions of the GNU General Public License can be used, that proxy's public statement of acceptance of a version permanently authorizes you to choose that version for the Program. Later license versions may give you additional or different permissions. However, no additional obligations are imposed on any author or copyright holder as a result of your choosing to follow a later version. 15. Disclaimer of Warranty. THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 16. Limitation of Liability. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. 17. Interpretation of Sections 15 and 16. If the disclaimer of warranty and limitation of liability provided above cannot be given local legal effect according to their terms, reviewing courts shall apply local law that most closely approximates an absolute waiver of all civil liability in connection with the Program, unless a warranty or assumption of liability accompanies a copy of the Program in return for a fee. END OF TERMS AND CONDITIONS How to Apply These Terms to Your New Programs If you develop a new program, and you want it to be of the greatest possible use to the public, the best way to achieve this is to make it free software which everyone can redistribute and change under these terms. To do so, attach the following notices to the program. It is safest to attach them to the start of each source file to most effectively state the exclusion of warranty; and each file should have at least the "copyright" line and a pointer to where the full notice is found. Copyright (C) This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . Also add information on how to contact you by electronic and paper mail. If the program does terminal interaction, make it output a short notice like this when it starts in an interactive mode: Copyright (C) This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. This is free software, and you are welcome to redistribute it under certain conditions; type `show c' for details. The hypothetical commands `show w' and `show c' should show the appropriate parts of the General Public License. Of course, your program's commands might be different; for a GUI interface, you would use an "about box". You should also get your employer (if you work as a programmer) or school, if any, to sign a "copyright disclaimer" for the program, if necessary. For more information on this, and how to apply and follow the GNU GPL, see . The GNU General Public License does not permit incorporating your program into proprietary programs. If your program is a subroutine library, you may consider it more useful to permit linking proprietary applications with the library. If this is what you want to do, use the GNU Lesser General Public License instead of this License. But first, please read .