Fix mmCIF parsing bug with underscores

* Allow underscore in non-first columns
* Add test case

Squashed commit of pull request #1349

Bio/PDB/MMCIF2Dict.py

@@ -38,7 +38,11 @@ class MMCIF2Dict(dict):
                     n = 0
                     continue
                 elif loop_flag:
-                    if token.startswith("_"):
+                    # The second condition checks we are in the first column
+                    # Some mmCIF files (e.g. 4q9r) have values in later columns
+                    # starting with an underscore and we don't want to read
+                    # these as keys
+                    if token.startswith("_") and (n == 0 or i % n == 0):
                         if i > 0:
                             loop_flag = False
                         else:

Tests/PDB/4Q9R_min.cif

@@ -0,0 +1,10 @@
+data_4Q9R
+loop_
+_pdbx_audit_revision_item.ordinal
+_pdbx_audit_revision_item.revision_ordinal
+_pdbx_audit_revision_item.data_content_type
+_pdbx_audit_revision_item.item
+1  5 'Structure model' '_atom_site.B_iso_or_equiv'
+2  5 'Structure model' '_atom_site.Cartn_x'
+3  5 'Structure model' '_atom_site.Cartn_y'
+4  5 'Structure model' '_atom_site.Cartn_z'