drop pandas requirement?

[speleo3]

pdb2pqr uses pandas in only one place for a pretty trivial task: Pass a table from run_propka() to non_trivial() only to extract two columns. The code could easily be changed to:

--- a/pdb2pqr/main.py
+++ b/pdb2pqr/main.py
@@ -561,8 +560,7 @@ def run_propka(args, biomolecule):
         else:
             row_dict["coupled_group"] = None
         rows.append(row_dict)
-    df = pandas.DataFrame(rows)
-    return df, pka_str
+    return rows, pka_str
 
 
 def non_trivial(args, biomolecule, ligand, definition, is_cif):
@@ -621,7 +619,7 @@ def non_trivial(args, biomolecule, ligand, definition, is_cif):
             biomolecule.apply_pka_values(
                 forcefield_.name,
                 args.ph,
-                dict(zip(pka_df.group_label, pka_df.pKa)),
+                dict((row['group_label'], row['pKa']) for row in pka_df),
             )
         _LOGGER.info("Adding hydrogens to biomolecule.")
         biomolecule.add_hydrogens()

With this change, we could drop pandas from install_requires and move it to tests_require (it's used for a few tests).

I like to keep run dependencies lightweight when possible, and dropping pandas would mean dropping 10+MB when packaging pdb2pqr with PyMOL.

If run_propka() is considered part of the public API, then this would be a breaking change. Although compatible code could be written like this:

pka_list_or_df, pka_str = run_propka(args, biomolecule)
pka_df = pandas.DataFrame(pka_list_or_df)

[sobolevnrm]

That makes sense to me. I can try to work on it this weekend unless you or @intendo want to submit a PR sooner. Thanks!

[speleo3]

I already had the patch ready to be uploaded :-)
PR #180

[sobolevnrm]

Nice; thank you!