fix(--gtdbtk_classification_file): Add method to parse taxonomy from …

…classify file.

Closes #428

gtdbtk/files/classify_summary.py

@@ -111,6 +111,15 @@ def has_row(self) -> bool:
             return True
         return False
 
+    def get_gid_taxonomy(self) -> Dict[str, List[str]]:
+        out = dict()
+        for gid, row in self.rows.items():
+            split_tax = row.classification.split(';')
+            if len(split_tax) != 7:
+                raise GTDBTkExit(f'Expected a 7-rank taxonomy for {gid} but got {row.classification}')
+            out[gid] = split_tax
+        return out
+
     def write(self):
         """Writes the summary file to disk. None will be replaced with N/A"""
         with open(self.path, 'w') as fh:
@@ -174,5 +183,3 @@ class ClassifySummaryFileBAC120(ClassifySummaryFile):
     def __init__(self, out_dir: str, prefix: str):
         path = os.path.join(out_dir, PATH_BAC120_SUMMARY_OUT.format(prefix=prefix))
         super().__init__(path, 'bac120')
-
-

gtdbtk/main.py

@@ -41,7 +41,7 @@
 from gtdbtk.external.fasttree import FastTree
 from gtdbtk.infer_ranks import InferRanks
 from gtdbtk.files.batchfile import Batchfile
-from gtdbtk.files.classify_summary import ClassifySummaryFileAR53
+from gtdbtk.files.classify_summary import ClassifySummaryFileAR53, ClassifySummaryFile
 from gtdbtk.markers import Markers
 from gtdbtk.misc import Misc
 from gtdbtk.model.enum import Domain
@@ -204,16 +204,24 @@ def _read_taxonomy_files(self, options) -> Dict[str, Tuple[str, str, str, str, s
             check_file_exists(options.gtdbtk_classification_file)
 
             self.logger.info('Reading GTDB-Tk classification file.')
-            gtdbtk_taxonomy = Taxonomy().read(options.gtdbtk_classification_file)
-            del gtdbtk_taxonomy['user_genome']
-            num_reassigned = 0
+            gtdbtk_classify_file = ClassifySummaryFile(path=options.gtdbtk_classification_file)
+            gtdbtk_classify_file.read()
+            gtdbtk_taxonomy = gtdbtk_classify_file.get_gid_taxonomy()
+            if len(gtdbtk_taxonomy) == 0:
+                raise GTDBTkExit(f'No genomes found in GTDB-Tk classification file: {options.gtdbtk_classification_file}')
+
+            num_rep_reassigned = 0
+            num_usr_reassigned = 0
             for gid, taxa in gtdbtk_taxonomy.items():
                 if gid in taxonomy:
-                    num_reassigned += 1
+                    num_rep_reassigned += 1
+                else:
+                    num_usr_reassigned += 1
                 taxonomy[gid] = taxa
 
             self.logger.info(f'Read GTDB-Tk classifications for {len(gtdbtk_taxonomy):,} genomes.')
-            self.logger.info(f'Reassigned taxonomy for {num_reassigned:,} GTDB representative genomes.')
+            self.logger.info(f'Reassigned taxonomy for {num_rep_reassigned:,} GTDB representative '
+                             f'genomes, and {num_usr_reassigned:,} query genomes.')
 
         if options.custom_taxonomy_file:
             # add and overwrite taxonomy for genomes specified in the