compute cliques

scripts/EPPO-match.py

 import taxutils
 import collections
 import os
+import os.path
 
 
 eppo_taxonomy = taxutils.EPPOParser.read_xmlfull_file('output/EPPO/fullcodes.xml')
@@ -50,11 +51,17 @@ for eppo_taxon in eppo_taxonomy.taxa():
                     mapping.add_alignment(ncbi_taxon, eppo_taxon, 'with-ancestor')
 
 
-os.makedirs('output/EPPO-match', exist_ok=True)
+OUTDIR = os.path.join('output', 'EPPO-match')
+os.makedirs(OUTDIR, exist_ok=True)
 
 
-taxutils.Logger.info('writing output/EPPO-match/matched-codes.csv')
-with open('output/EPPO-match/matched-codes.csv', 'w') as f:
+def outfile(filename):
+    fullname = os.path.join(OUTDIR, filename)
+    taxutils.Logger.info('writing %s' % fullname)
+    return open(fullname, 'w')
+
+
+with outfile('matched-codes.csv') as f:
     f.write('\t'.join((
         'EPPO code',
         'EPPO name',
@@ -87,8 +94,7 @@ with open('output/EPPO-match/matched-codes.csv', 'w') as f:
                 f.write('\n')
 
 
-taxutils.Logger.info('writing output/EPPO-match/unmatched-codes.csv')
-with open('output/EPPO-match/unmatched-codes.csv', 'w') as f:
+with outfile('unmatched-codes.csv') as f:
     f.write('\t'.join((
         'EPPO code',
         'EPPO name',
@@ -105,4 +111,55 @@ with open('output/EPPO-match/unmatched-codes.csv', 'w') as f:
             f.write('\n')
 
 
+CLIQUES = {}
+CLIQUE_MAP = {}
+for next_clique, eppo_taxon in enumerate(eppo_taxonomy.taxa()):
+    if eppo_taxon.taxid in CLIQUE_MAP:
+        eppo_clique = CLIQUE_MAP[eppo_taxon.taxid]
+        eppo_clique_ncbi_taxa, eppo_clique_eppo_taxa = CLIQUES[eppo_clique]
+    else:
+        eppo_clique = next_clique
+        eppo_clique_ncbi_taxa = []
+        eppo_clique_eppo_taxa = [eppo_taxon]
+        CLIQUES[eppo_clique] = (eppo_clique_ncbi_taxa, eppo_clique_eppo_taxa)
+        CLIQUE_MAP[eppo_taxon.taxid] = eppo_clique
+    if mapping.has_21(eppo_taxon):
+        for ncbi_taxon, alignment_type in mapping.get_21(eppo_taxon):
+            if ncbi_taxon.taxid in CLIQUE_MAP:
+                ncbi_clique = CLIQUE_MAP[ncbi_taxon.taxid]
+                if ncbi_clique == eppo_clique:
+                    continue
+                (ncbi_clique_ncbi_taxa, ncbi_clique_eppo_taxa) = CLIQUES[ncbi_clique]
+                eppo_clique_ncbi_taxa.extend(ncbi_clique_ncbi_taxa)
+                eppo_clique_eppo_taxa.extend(ncbi_clique_eppo_taxa)
+                for t in ncbi_clique_ncbi_taxa:
+                    CLIQUE_MAP[t.taxid] = eppo_clique
+                for t in ncbi_clique_eppo_taxa:
+                    CLIQUE_MAP[t.taxid] = eppo_clique
+                del CLIQUES[ncbi_clique]
+            else:
+                CLIQUE_MAP[ncbi_taxon.taxid] = eppo_clique
+                eppo_clique_ncbi_taxa.append(ncbi_taxon)
+
+
+eppo_max = 0
+ncbi_max = 0
+both_max = 0
+for ncbi_taxa, eppo_taxa in CLIQUES.values():
+    eppo_max = max(eppo_max, len(eppo_taxa))
+    ncbi_max = max(ncbi_max, len(ncbi_taxa))
+    both_max = max(both_max, len(eppo_taxa) + len(ncbi_taxa))
+    if len(eppo_taxa) == 1:
+        if len(ncbi_taxa) < 2:
+            continue
+        prefix = '>'
+    elif len(ncbi_taxa) == 1:
+        prefix = '<'
+    else:
+        prefix = '*'
+    print('%s %s = %s' % (prefix, ', '.join(t.taxid for t in eppo_taxa), ', '.join(t.taxid for t in ncbi_taxa)))
+print('eppo_max = %d' % eppo_max)
+print('ncbi_max = %d' % ncbi_max)
+print('both_max = %d' % both_max)
+
 taxutils.Logger.info('Done')