fix: Add support for numeric and precision with length and precision in Postgres Custom Query

[sharangagarwal]

Closes #681

• Added _parse_numeric function from the new ibis postgres datatype update and made it compatible with the current version
• added parsy pypi library dependency

[sharangagarwal]

/gcbrun

[sharangagarwal]

/gcbrun

[sharangagarwal]

/gcbrun

[nj1973]

Have you tested your code with non-numeric columns? I tried it with a date and got an error:

  File "/usr/local/google/home/neiljohnson/github/professional-services-data-validator2/third_party/ibis/ibis_postgres/client.py", line 107, in _parse_numeric
    return ty.parse(text)
  File "/usr/local/google/home/neiljohnson/github/professional-services-data-validator2/.venv/lib/python3.10/site-packages/parsy/__init__.py", line 87, in parse
    (result, _) = (self << eof).parse_partial(stream)
  File "/usr/local/google/home/neiljohnson/github/professional-services-data-validator2/.venv/lib/python3.10/site-packages/parsy/__init__.py", line 101, in parse_partial
    raise ParseError(result.expected, stream, result.furthest)
parsy.ParseError: expected one of 'decimal', 'numeric' at 0:0

I think you should add a unit test feeding each common data type string into the _get_type function to be sure we don't get any exceptions.

[nj1973]

The test column triggering the exception is actually timestamp rather than date.

col_ts          timestamp(6)

And looks like this from a describe:

    Column    |              Type              | Collation | Nullable | Default
--------------+--------------------------------+-----------+----------+---------
 id           | integer                        |           | not null |
...
 col_ts       | timestamp(6) without time zone |           |          |

[sharangagarwal]

/gcbrun

[nj1973]

LGTM