Added tabular format indexer and related tests

Bio/SearchIO/BlastIO/blast_tab.py

@@ -545,7 +545,7 @@ class BlastTabIndexer(SearchIndexer):
                     "Custom fields is missing an ID column. One of these must be "
                     "present: 'qseqid', 'qacc', or 'qaccver'."
                 )
-
+                
     def __iter__(self):
         """Iterate over the file handle; yields key, start offset, and length."""
         handle = self._handle

Bio/SearchIO/InfernalIO/__init__.py

@@ -7,6 +7,7 @@
 
 
 from .infernal_tab import InfernalTabParser
+from .infernal_tab import InfernalTabIndexer
 from .infernal_text import InfernalTextParser
 
 

Bio/SearchIO/InfernalIO/_base.py

@@ -43,3 +43,4 @@ class _BaseInfernalParser:
                 setattr(hit, attr, value)
             hit_list.append(hit)
         return hit_list
+

Bio/SearchIO/InfernalIO/infernal_tab.py

@@ -17,7 +17,7 @@ from Bio.SearchIO.HmmerIO import Hmmer3TabParser
 from ._base import _BaseInfernalParser
 
 
-__all__ = ("InfernalTabParser")
+__all__ = ("InfernalTabParser", "InfernalTabIndexer")
 
 
 # tabular format column names
@@ -251,6 +251,97 @@ class InfernalTabParser(_BaseInfernalParser):
             self.line = self.handle.readline()
 
 
+class InfernalTabIndexer(SearchIndexer):
+    """Indexer class for Infernal tabular output."""
+
+
+    _parser = InfernalTabParser
+
+
+    def __iter__(self):
+        """Iterate over the file handle; yields key, start offset, and length."""
+        handle = self._handle
+        handle.seek(0)
+        qresult_key = None
+        query_id_idx = None
+        # mark for the comments
+        comment_mark = b"#"
+        # mark for the row names line
+        row_names_mark = b"#target name"
+        row_names_mark_fmt_2 = b"#idx"
+        split_mark = b" "
+
+        # set line with initial mock value, to emulate header
+        line = comment_mark
+        
+        # read through header and get the query index position based on the
+        # tabular output format
+        while line.startswith(comment_mark):
+            start_offset = handle.tell()
+            line = handle.readline()
+            if line.startswith(row_names_mark):
+                query_id_idx = 2
+            elif line.startswith(row_names_mark_fmt_2):
+                query_id_idx = 3
+
+        assert query_id_idx is not None
+
+        # handle empty file
+        if not line:
+            return
+
+        # index the qresults
+        while True:
+            # get end offset here since we only know a qresult ends after
+            # encountering the next one
+            end_offset = handle.tell()
+            
+            # process the previous line
+            cols = [x for x in line.strip().split(split_mark) if x]
+            if qresult_key is None:
+                qresult_key = cols[query_id_idx]
+            else:
+                curr_key = cols[query_id_idx]
+                if curr_key != qresult_key:
+                    adj_end = end_offset - len(line)
+                    yield (qresult_key.decode(), start_offset, adj_end - start_offset)
+                    qresult_key = curr_key
+                    start_offset = adj_end
+
+            line = handle.readline()
+            if not line or line.startswith(comment_mark):
+                yield (qresult_key.decode(), start_offset, end_offset - start_offset)
+                break
+
+
+    def get_raw(self, offset):
+        """Return the raw bytes string of a QueryResult object from the given offset."""
+        handle = self._handle
+        handle.seek(offset)
+        query_id_idx = 2
+        qresult_key = None
+        qresult_raw = b""
+        split_mark = b" "
+        comment_mark = b"#"
+
+        while True:
+            line = handle.readline()
+            if not line or line.startswith(comment_mark):
+                break
+            cols = [x for x in line.strip().split(split_mark) if x]
+            if qresult_key is None:
+                qresult_key = cols[query_id_idx]
+            else:
+                curr_key = cols[query_id_idx]
+                if curr_key != qresult_key:
+                    break
+            qresult_raw += line
+
+
+
+        return qresult_raw
+
+
 # if not used as a module, run the doctest
 if __name__ == "__main__":
     from Bio._utils import run_doctest

Bio/SearchIO/__init__.py

@@ -245,7 +245,6 @@ _INDEXER_MAP = {
     "hmmscan3-domtab": ("HmmerIO", "Hmmer3DomtabHmmhitIndexer"),
     "hmmsearch3-domtab": ("HmmerIO", "Hmmer3DomtabHmmqueryIndexer"),
     "infernal-tab": ("InfernalIO", "InfernalTabIndexer"),
-    "infernal-text": ("InfernalIO", "InfernalTextIndexer"),
     "phmmer3-domtab": ("HmmerIO", "Hmmer3DomtabHmmqueryIndexer"),
 }
 

Tests/test_SearchIO_infernal_tab_index.py

@@ -18,59 +18,58 @@ from search_tests_common import CheckRaw
 class InfernalTabRawCases(CheckRaw):
     fmt = "infernal-tab"
 
+
     def test_infernal_tab_single(self):
-        """Test infernal-tab raw string retrieval, cmsearch, single query (U2_Yeast)."""
+        """Test infernal-tab raw string retrieval, cmsearch, single query."""
-        filename = os.path.join("Infernal", "cmsearch_114_U2_Yeast.tbl")
+        filename = os.path.join("Infernal", "U2_Yeast-threshold.tbl")
-        raw = """#target name            accession query name           accession mdl mdl from   mdl to seq from   seq to strand trunc pass   gc  bias  score   E-value inc description of target
+        raw = """ENA|BK006936|BK006936.2 -         U2                   RF00004    cm        1      193   681858   681747      -    no    1 0.33   0.1   98.7   5.9e-20 !   TPA_inf: Saccharomyces cerevisiae S288C chromosome II, complete sequence.
-#---------------------- --------- -------------------- --------- --- -------- -------- -------- -------- ------ ----- ---- ---- ----- ------ --------- --- ---------------------
-ENA|BK006936|BK006936.2 -         U2                   RF00004    cm        1      193   681858   681747      -    no    1 0.33   0.1   98.7   5.9e-20 !   TPA_inf: Saccharomyces cerevisiae S288C chromosome II, complete sequence.
 """
         self.check_raw(filename, "U2", raw)
 
-    def test_infernal_tab_multiple_first(self):
+    
+    def test_infernal_tab_multiple(self):
+        """Test infernal-tab raw string retrieval, cmsearch, single query, multiple non-consecutive hits."""
+        filename = os.path.join("Infernal", "U2_Yeast-shuf.tbl")
+        raw = """ENA|BK006936|BK006936.2 -         U2                   RF00004    cm        1      193   681858   681747      -    no    1 0.33   0.1   98.7   1.3e-20 !   -
+ENA|BK006948|BK006948.2 -         U2                   RF00004    cm        1      193   737498   737324      -    no    1 0.39   0.0   19.8      0.11 ?   -
+ENA|BK006936|BK006936.2 -         U2                   RF00004    cm        1      193  1370418  1370563      +    no    1 0.34   0.1   15.6       1.1 ?   -
+ENA|BK006936|BK006936.2 -         U2                   RF00004    cm        1      193  1079243  1079392      +    no    1 0.39   0.0   15.3       1.3 ?   -
+ENA|BK006948|BK006948.2 -         U2                   RF00004    cm        1      193   425490   425693      +    no    1 0.34   0.9   13.7       3.1 ?   -
+ENA|BK006948|BK006948.2 -         U2                   RF00004    cm        1      193  1073786  1073950      +    no    1 0.33   0.5   11.9       8.3 ?   -
+"""
+        self.check_raw(filename, "U2", raw)
+
+
+    def test_infernal_tab_single_first(self):
         """Test infernal-tab raw string retrieval, cmsearch, multiple queries, first."""
-        filename = os.path.join("Infernal", "cmscan_115_IRES_5S_U2_Yeast.tbl")
+        filename = os.path.join("Infernal", "IRES_5S_U2_Yeast-cmscan.tbl")
-        raw = """#target name         accession query name              accession mdl mdl from   mdl to seq from   seq to strand trunc pass   gc  bias  score   E-value inc description of target
+        raw = """U2                   RF00004   ENA|BK006935|BK006935.2 -          cm        1      193    52929    53083      +    no    1 0.44   0.0   13.5      0.91 ?   U2 spliceosomal RNA
-#------------------- --------- ----------------------- --------- --- -------- -------- -------- -------- ------ ----- ---- ---- ----- ------ --------- --- ---------------------
-U2                   RF00004   ENA|BK006935|BK006935.2 -          cm        1      193    52929    53083      +    no    1 0.44   0.0   13.5      0.91 ?   U2 spliceosomal RNA
 U2                   RF00004   ENA|BK006935|BK006935.2 -          cm        1      193   196571   196389      -    no    1 0.33   5.3   12.8       1.3 ?   U2 spliceosomal RNA
 """
         self.check_raw(filename, "ENA|BK006935|BK006935.2", raw)
 
+    
     def test_infernal_tab_multiple_middle(self):
         """Test infernal-tab raw string retrieval, cmsearch, multiple queries, middle."""
-        filename = os.path.join("Infernal", "cmscan_115_IRES_5S_U2_Yeast.tbl")
+        filename = os.path.join("Infernal", "IRES_5S_U2_Yeast-cmscan.tbl")
-        raw = """#target name         accession query name              accession mdl mdl from   mdl to seq from   seq to strand trunc pass   gc  bias  score   E-value inc description of target
+        raw = """U2                   RF00004   ENA|BK006936|BK006936.2 -          cm        1      193   681858   681747      -    no    1 0.33   0.1   98.7   1.2e-20 !   U2 spliceosomal RNA
-#------------------- --------- ----------------------- --------- --- -------- -------- -------- -------- ------ ----- ---- ---- ----- ------ --------- --- ---------------------
-U2                   RF00004   ENA|BK006936|BK006936.2 -          cm        1      193   681858   681747      -    no    1 0.33   0.1   98.7   1.2e-20 !   U2 spliceosomal RNA
 """
         self.check_raw(filename, "ENA|BK006936|BK006936.2", raw)
 
+
     def test_infernal_tab_multiple_last(self):
         """Test infernal-tab raw string retrieval, cmsearch, multiple queries, last."""
-        filename = os.path.join("Infernal", "cmscan_115_IRES_5S_U2_Yeast.tbl")
+        filename = os.path.join("Infernal", "IRES_5S_U2_Yeast-cmscan.tbl")
-        raw = """#target name         accession query name              accession mdl mdl from   mdl to seq from   seq to strand trunc pass   gc  bias  score   E-value inc description of target
+        raw = """5S_rRNA              RF00001   ENA|BK006937|BK006937.2 -          cm        1      119      761      644      -    no    1 0.41   0.3   14.1       2.4 ?   5S ribosomal RNA
-#------------------- --------- ----------------------- --------- --- -------- -------- -------- -------- ------ ----- ---- ---- ----- ------ --------- --- ---------------------
-5S_rRNA              RF00001   ENA|BK006937|BK006937.2 -          cm        1      119      761      644      -    no    1 0.41   0.3   14.1       2.4 ?   5S ribosomal RNA
 U2                   RF00004   ENA|BK006937|BK006937.2 -          cm        1      193   229986   229885      -    no    1 0.32   0.1   11.1       4.7 ?   U2 spliceosomal RNA
 """
         self.check_raw(filename, "ENA|BK006937|BK006937.2", raw)
 
+    
     def test_infernal_tab_multiple_fmt_2(self):
         """Test infernal-tab raw string retrieval, cmsearch, multiple queries, fmt 2."""
-        filename = os.path.join("Infernal", "cmscan_115_IRES_5S_U2_Yeast_fmt_2.tbl")
+        filename = os.path.join("Infernal", "IRES_5S_U2_Yeast-cmscan-fmt_2.tbl")
-        raw = """#idx target name          accession query name              accession clan name mdl mdl from   mdl to seq from   seq to strand trunc pass   gc  bias  score   E-value inc olp anyidx afrct1 afrct2 winidx wfrct1 wfrct2 mdl len seq len description of target
+        raw = """1    U2                   RF00004   ENA|BK006936|BK006936.2 -         -          cm        1      193   681858   681747      -    no    1 0.33   0.1   98.7   1.2e-20  !   *       -      -      -      -      -      -     193  813184 U2 spliceosomal RNA
-#--- -------------------- --------- ----------------------- --------- --------- --- -------- -------- -------- -------- ------ ----- ---- ---- ----- ------ --------- --- --- ------ ------ ------ ------ ------ ------ ------- ------- ---------------------
-1    U2                   RF00004   ENA|BK006936|BK006936.2 -         -          cm        1      193   681858   681747      -    no    1 0.33   0.1   98.7   1.2e-20  !   *       -      -      -      -      -      -     193  813184 U2 spliceosomal RNA
-"""
-        self.check_raw(filename, "ENA|BK006936|BK006936.2", raw, fmt=2)
-
-    def test_infernal_tab_multiple_fmt_2_infer(self):
-        """Test infernal-tab raw string retrieval, cmsearch, multiple queries, fmt 2, inferred."""
-        filename = os.path.join("Infernal", "cmscan_115_IRES_5S_U2_Yeast_fmt_2.tbl")
-        raw = """#idx target name          accession query name              accession clan name mdl mdl from   mdl to seq from   seq to strand trunc pass   gc  bias  score   E-value inc olp anyidx afrct1 afrct2 winidx wfrct1 wfrct2 mdl len seq len description of target
-#--- -------------------- --------- ----------------------- --------- --------- --- -------- -------- -------- -------- ------ ----- ---- ---- ----- ------ --------- --- --- ------ ------ ------ ------ ------ ------ ------- ------- ---------------------
-1    U2                   RF00004   ENA|BK006936|BK006936.2 -         -          cm        1      193   681858   681747      -    no    1 0.33   0.1   98.7   1.2e-20  !   *       -      -      -      -      -      -     193  813184 U2 spliceosomal RNA
 """
         self.check_raw(filename, "ENA|BK006936|BK006936.2", raw)
 
@@ -78,24 +77,28 @@ U2                   RF00004   ENA|BK006937|BK006937.2 -          cm        1
 class InfernalTabIndexCases(CheckIndex):
     fmt = "infernal-tab"
 
-    def test_infernal_tab_1q(self):
+
-        """Test infernal-tab indexing, cmsearch, one query, one hit."""
+    def test_infernal_tab_single(self):
-        filename = os.path.join("Infernal", "cmsearch_114_U2_Yeast.tbl")
+        """Test infernal-tab indexing, cmsearch, single query."""
+        filename = os.path.join("Infernal", "U2_Yeast-threshold.tbl")
         self.check_index(filename, self.fmt)
 
-    def test_infernal_tab_1q_0m(self):
+    
-        """Test infernal-tab indexing, cmsearch, single query, no hits."""
+    def test_infernal_tab_single_no_hit(self):
-        filename = os.path.join("Infernal", "cmsearch_114_IRES_Yeast.tbl")
+        """Test infernal-tab indexing, cmsearch, single query."""
+        filename = os.path.join("Infernal", "IRES_Yeast.tbl")
         self.check_index(filename, self.fmt)
 
-    def test_infernal_tab_1q_mm(self):
-        """Test infernal-tab indexing, cmsearch, single query, multiple hits."""
-        filename = os.path.join("Infernal", "cmsearch_114_5S_Yeast.tbl")
-        self.check_index(filename, self.fmt)
 
-    def test_infernal_tab_mq_mm(self):
+    def test_infernal_tab_single_multiple_hit(self):
-        """Test infernal-tab indexing, cmscan, multiple query, multiple matches."""
+        """Test infernal-tab indexing, cmsearch, single query."""
-        filename = os.path.join("Infernal", "cmscan_115_IRES_5S_U2_Yeast.tbl")
+        filename = os.path.join("Infernal", "5S_Yeast.tbl")
+        self.check_index(filename, self.fmt)
+    
+
+    def test_infernal_tab_multiple_multiple_hit(self):
+        """Test infernal-tab indexing, cmscan, multiple query."""
+        filename = os.path.join("Infernal", "IRES_5S_U2_Yeast-cmscan.tbl")
         self.check_index(filename, self.fmt)