kkonganti@0: """Pipeline script"""
kkonganti@0: 
kkonganti@0: import csv, datetime, logging, re, sqlite3, sys
kkonganti@0: import lexmapr.create_databases as cdkp
kkonganti@0: import lexmapr.ontology_reasoner as ontr
kkonganti@0: import lexmapr.pipeline_helpers as helpers
kkonganti@0: import lexmapr.pipeline_resources as pipeline_resources
kkonganti@0: import lexmapr.run_summary as summarize
kkonganti@0: from itertools import permutations
kkonganti@0: from collections import OrderedDict
kkonganti@0: from nltk.tokenize import word_tokenize
kkonganti@0: from lexmapr.definitions import not_provided, arg_bins, ontol_db, ontol_interest
kkonganti@0: 
kkonganti@0: 
kkonganti@0: # TODO: deal with ambiguous terms: nut milk, dairy (building ENVO_00003862 v dairy food product)
kkonganti@0: # TODO: combine ScorLex.csv and misspellings.csv and edit _add_predefined_resources function
kkonganti@0: # TODO: decide what to do with same synonym, ex custom-customize vs custom-tradition
kkonganti@0: # TODO: make web on database instead of pulling relationships from API?
kkonganti@0: # TODO: remove synonyms over time from SynLex and resource_synonyms.csv; edit get_synonyms
kkonganti@0: 
kkonganti@0: def run(run_args):
kkonganti@0:     '''Main text processing and mapping pipeline'''
kkonganti@0: 
kkonganti@0:     # Add information from EMBL and predefined_resources folder
kkonganti@0:     t0 = datetime.datetime.now()
kkonganti@0:     global ontol_interest
kkonganti@0:     if run_args.embl_ontol:
kkonganti@0:         ontol_interest = run_args.embl_ontol
kkonganti@0: 
kkonganti@0:     print('\nBuilding databases...')
kkonganti@0:     cdkp.get_synonyms(run_args.remake_cache, ontol_interest)
kkonganti@0:     cdkp.get_resource_ids(run_args.remake_cache, ontol_interest)
kkonganti@0:     lookup_table = pipeline_resources.get_predefined_resources()
kkonganti@0:     t1 = datetime.datetime.now()
kkonganti@0:     print(f'\tDone! {t1-t0} passed'.ljust(60) + '\nMapping terms...')
kkonganti@0:     logging.info(f'Database build/confirm: {t1}')
kkonganti@0: 
kkonganti@0:     # Apply other arguments and initiate mapping cache
kkonganti@0:     term_cache = {'':'\t\t\t\t',}
kkonganti@0:     output_fields = ['Sample_Id',
kkonganti@0:                      'Sample_Desc',
kkonganti@0:                      'Processed_Sample',
kkonganti@0:                      'Annotated_Sample',
kkonganti@0:                      'Matched_Components']
kkonganti@0:     if run_args.full:
kkonganti@0:         output_fields += ['Match_Status (Macro Level)',
kkonganti@0:                           'Match_Status (Micro Level)',
kkonganti@0:                           'Sample_Transformations']
kkonganti@0:         term_cache[''] += '\t\t'
kkonganti@0:     else:
kkonganti@0:         output_fields += ['Match_Status (Macro Level)']
kkonganti@0: 
kkonganti@0:     if run_args.bin:
kkonganti@0:         global arg_bins
kkonganti@0:         if run_args.user_bin is not None:
kkonganti@0:             arg_bins = run_args.user_bin
kkonganti@0:         for x in arg_bins:
kkonganti@0:             arg_bins[x] = ontr.Ontology_package(x, arg_bins[x])
kkonganti@0:         term_cache[''] += '\t'*len(arg_bins)
kkonganti@0:         output_fields += list(arg_bins.keys())
kkonganti@0:     else:
kkonganti@0:         arg_bins = {}
kkonganti@0: 
kkonganti@0:     OUT_file = open(run_args.output, 'w') if run_args.output else sys.stdout
kkonganti@0:     if OUT_file is sys.stdout:
kkonganti@0:         OUT_file.write('\n')
kkonganti@0:     OUT_file.write('\t'.join(output_fields))
kkonganti@0: 
kkonganti@0:     IN_file = open(run_args.input, 'r')
kkonganti@0:     if run_args.input[-4:] == '.csv':
kkonganti@0:         fr_reader = csv.reader(IN_file, delimiter=',')
kkonganti@0:     elif run_args.input[-4:] == '.tsv':
kkonganti@0:         fr_reader = csv.reader(IN_file, delimiter='\t')
kkonganti@0:     next(fr_reader)
kkonganti@0:     
kkonganti@0:     # Connect to primary database
kkonganti@0:     conn = sqlite3.connect(ontol_db)
kkonganti@0:     c = conn.cursor()
kkonganti@0: 
kkonganti@0:     # Iterate over samples in input file
kkonganti@0:     for sample_row in fr_reader:
kkonganti@0:         sample_id = '\n' + sample_row[0].strip() + '\t' + ','.join(sample_row[1:])
kkonganti@0:         original_sample = ' '.join(sample_row[1:]).strip()
kkonganti@0:         cleaned_sample = ''
kkonganti@0:         cleaned_annotated = ''
kkonganti@0:         macro_status = 'No Match'
kkonganti@0:         matched_components = []
kkonganti@0:         synonym_match = []
kkonganti@0:         micro_status = []
kkonganti@0:         bin_class = {x:[] for x in arg_bins}
kkonganti@0:         ancestors = set()
kkonganti@0:         sample_conversion_status = {}
kkonganti@0:         synonym_map = {}
kkonganti@0:         treated_sample = helpers.punctuation_treatment(original_sample)
kkonganti@0: 
kkonganti@0:         # Determine if sample in predefined list of null values
kkonganti@0:         if treated_sample in not_provided:
kkonganti@0:             write_line = '\t' + treated_sample + '\tNot annotated\t\t' + macro_status
kkonganti@0:             OUT_file.write(sample_id + write_line)
kkonganti@0:             if run_args.full:
kkonganti@0:                 OUT_file.write('\t' + str(micro_status) + '\t' + str(sample_conversion_status))
kkonganti@0:             if run_args.bin:
kkonganti@0:                 OUT_file.write('\t'*len(bin_class))
kkonganti@0:             continue
kkonganti@0: 
kkonganti@0:         # Remove negated words and some compound words, apply corrections
kkonganti@0:         proc_sample = helpers.further_cleanup(treated_sample)
kkonganti@0:         proc_sample, micro_status = helpers.process_sample(proc_sample,lookup_table,micro_status)
kkonganti@0: 
kkonganti@0:         # Try finding processed sample in cache
kkonganti@0:         try:
kkonganti@0:             OUT_file.write(sample_id+term_cache[proc_sample])
kkonganti@0:             continue
kkonganti@0:         except(KeyError):
kkonganti@0:             pass
kkonganti@0: 
kkonganti@0:         # Attempt full term matches with and without suffixes
kkonganti@0:         if OUT_file is not sys.stdout:
kkonganti@0:             print('\tMatching '+sample_row[0].strip()+' '+ \
kkonganti@0:                                 '{:<40}'.format(proc_sample[:40]).ljust(60), end='\r')
kkonganti@0: 
kkonganti@0:         full_term_match = helpers.map_term(treated_sample, lookup_table, c)
kkonganti@0:         if full_term_match == []:
kkonganti@0:             full_term_match = helpers.map_term(proc_sample, lookup_table, c)
kkonganti@0:             if full_term_match != []:
kkonganti@0:                 micro_status.insert(0, 'Used Processed Sample')
kkonganti@0:         if full_term_match == [] and 'FOODON' in ontol_interest:
kkonganti@0:             full_term_match = helpers.map_term(proc_sample, lookup_table, c, True)
kkonganti@0:             if full_term_match != []:
kkonganti@0:                 micro_status.insert(0, 'Used Processed Sample')
kkonganti@0: 
kkonganti@0:         # Attempt full term match with cleaned sample using suffixes
kkonganti@0:         if full_term_match == []:
kkonganti@0:             for sw_token in word_tokenize(proc_sample):
kkonganti@0:                 if helpers.is_date(sw_token) or helpers.is_number(sw_token):
kkonganti@0:                     continue
kkonganti@0:                 lemma, micro_status = helpers.singularize_token(sw_token, lookup_table,
kkonganti@0:                                                                micro_status, c)
kkonganti@0:                 if not sw_token == lemma:
kkonganti@0:                     sample_conversion_status[sw_token] = lemma
kkonganti@0:                 cleaned_sample = helpers.get_cleaned_sample(cleaned_sample, lemma, lookup_table)
kkonganti@0:                 # Not de-duplicating tokens because can't account for all legitimate double names
kkonganti@0: 
kkonganti@0:             full_term_match = helpers.map_term(cleaned_sample, lookup_table, c)
kkonganti@0:             if full_term_match == [] and 'FOODON' in ontol_interest:
kkonganti@0:                 full_term_match = helpers.map_term(cleaned_sample, lookup_table, c, True)
kkonganti@0:             if full_term_match != []:
kkonganti@0:                 micro_status.insert(0, 'Used Cleaned Sample')
kkonganti@0: 
kkonganti@0:         # Combine the matched terms
kkonganti@0:         if full_term_match != []:
kkonganti@0:             for x in full_term_match:
kkonganti@0:                 matched_components.append(x['term'] + ':' + x['id'])
kkonganti@0:                 macro_status = 'Full Term Match'
kkonganti@0:                 micro_status += x['status']
kkonganti@0: 
kkonganti@0:         # Try matching permutations if full term match fails
kkonganti@0:         # Functions mostly retained from v 0.7
kkonganti@0:         if macro_status == 'No Match':
kkonganti@0:             covered_tokens = set()
kkonganti@0:             for i in range(5, 0, -1):
kkonganti@0:                 for gram_chunk in helpers.get_gram_chunks(cleaned_sample, i):
kkonganti@0:                     concat_gram_chunk = ' '.join(gram_chunk)
kkonganti@0:                     gram_permutations =\
kkonganti@0:                         list(OrderedDict.fromkeys(permutations(concat_gram_chunk.split())))
kkonganti@0:                     if set(gram_chunk) <= covered_tokens:
kkonganti@0:                         continue
kkonganti@0:                     for gram_permutation in gram_permutations:
kkonganti@0:                         gram_permutation_str = ' '.join(gram_permutation)
kkonganti@0:                         component_match = helpers.map_term(gram_permutation_str, lookup_table, c)
kkonganti@0:                         if not component_match and 'FOODON' in ontol_interest:
kkonganti@0:                             component_match = helpers.map_term(gram_permutation_str,
kkonganti@0:                                                          lookup_table, c, True)
kkonganti@0:                         if component_match:
kkonganti@0:                             for x in component_match:
kkonganti@0:                                 matched_components.append(x['term'] + ':' + x['id'])
kkonganti@0:                                 macro_status = 'Component Match'
kkonganti@0:                                 micro_status += x['status']
kkonganti@0:                                 covered_tokens.update(gram_chunk)
kkonganti@0: 
kkonganti@0:         # Try matching annotated synonyms if component match fails
kkonganti@0:         if macro_status == 'No Match':
kkonganti@0:             for clean_token in cleaned_sample.split():
kkonganti@0:                 cleaned_annotated,s_m=helpers.get_annotated_sample(cleaned_annotated,clean_token)
kkonganti@0:                 synonym_map.update(s_m)
kkonganti@0:         cleaned_annotated = helpers.annotation_reduce(cleaned_annotated, synonym_map)
kkonganti@0: 
kkonganti@0:         for x in helpers.get_annotated_synonyms(cleaned_annotated):
kkonganti@0:             synonym_match.extend(helpers.map_term(x, lookup_table, c))
kkonganti@0:         if synonym_match == [] and 'FOODON' in ontol_interest:
kkonganti@0:             for x in helpers.get_annotated_synonyms(cleaned_annotated):
kkonganti@0:                 synonym_match.extend(helpers.map_term(x, lookup_table, c, True))
kkonganti@0:         if synonym_match != []:
kkonganti@0:             macro_status = 'Synonym Match'
kkonganti@0:         for x in synonym_match:
kkonganti@0:             matched_components.append(x['term'] + ':' + x['id'])
kkonganti@0:             micro_status += x['status']
kkonganti@0: 
kkonganti@0:         # Remove matches that are ancestral to other matches
kkonganti@0:         if run_args.no_ancestors:
kkonganti@0:             for match_term in matched_components:
kkonganti@0:                 match_term = match_term.replace('NCBITAXON','NCBITaxon')
kkonganti@0:                 ontol_acc = ontr.Ontology_accession.make_instance(match_term)
kkonganti@0:                 ontol_anc = ontol_acc.get_family('ancestors')
kkonganti@0:                 try:
kkonganti@0:                     ontol_anc.remove('none found')
kkonganti@0:                 except(ValueError):
kkonganti@0:                     pass
kkonganti@0:                 ancestors |= set([x.id for x in ontol_anc])
kkonganti@0: 
kkonganti@0:             final_matches = []
kkonganti@0:             for match_term in matched_components:
kkonganti@0:                 if match_term.split(':')[-1].replace('NCBITAXON','NCBITaxon') not in ancestors:
kkonganti@0:                     final_matches.append(match_term)
kkonganti@0: 
kkonganti@0:         # Bin matches
kkonganti@0:         for x in arg_bins:
kkonganti@0:             for y in matched_components:
kkonganti@0:                 ontol_y = ontr.Ontology_accession.make_instance(y)
kkonganti@0:                 bin_class[x].extend(ontol_y.bin_term(arg_bins[x]))
kkonganti@0: 
kkonganti@0:         # Write to output
kkonganti@0:         if cleaned_annotated == '':
kkonganti@0:             cleaned_annotated = 'Not annotated'
kkonganti@0:         write_line = '\t' + cleaned_sample + '\t' + cleaned_annotated +\
kkonganti@0:                      '\t' + '|'.join(sorted(set(matched_components))) + '\t' + macro_status
kkonganti@0:         while re.search('  ', write_line):
kkonganti@0:             write_line = write_line.replace('  ',' ')
kkonganti@0:         term_cache[proc_sample] = write_line
kkonganti@0:         OUT_file.write(sample_id + write_line)
kkonganti@0:         if run_args.full:
kkonganti@0:             OUT_file.write('\t' + str(micro_status) + '\t' + str(sample_conversion_status))
kkonganti@0:             term_cache[proc_sample] += '\t'+str(micro_status)+'\t'+str(sample_conversion_status)
kkonganti@0:         if run_args.bin:
kkonganti@0:             for x in list(bin_class):
kkonganti@0:                 OUT_file.write('\t' + '|'.join(sorted(set(bin_class[x]))).replace('[]',''))
kkonganti@0:                 term_cache[proc_sample] += '\t' + '|'.join(sorted(set(bin_class[x])))
kkonganti@0: 
kkonganti@0: 
kkonganti@0:     IN_file.close()
kkonganti@0:     conn.close()
kkonganti@0:     if OUT_file is not sys.stdout:
kkonganti@0:         OUT_file.close()
kkonganti@0:     else:
kkonganti@0:         OUT_file.write('\n\n')
kkonganti@0: 
kkonganti@0:     # Report results to log and generate graphs
kkonganti@0:     t2 = datetime.datetime.now()
kkonganti@0:     print(f'\tDone! {t2-t1} passed'.ljust(60) + '\nReporting results...')
kkonganti@0:     if run_args.output:
kkonganti@0:         summarize.report_results(run_args.output, list(arg_bins.keys()))
kkonganti@0:         if run_args.graph == True:
kkonganti@0:             summarize.figure_folder()
kkonganti@0:             summarize.visualize_results(run_args.output, list(arg_bins.keys()))
kkonganti@0:     else:
kkonganti@0:         match_counts = summarize.report_cache(term_cache)
kkonganti@0:         if run_args.graph == True:
kkonganti@0:             summarize.figure_folder()
kkonganti@0:             summarize.visualize_cache(match_counts)
kkonganti@0: 
kkonganti@0:     print('\t'+f'Done! {datetime.datetime.now()-t2} passed'.ljust(60)+'\n')