feat: Allow implicit string → temporal conversion in SQL comparisons

[alexander-beedie]

Closes #15956.

Where available, the internal active schema is now associated with SQLExprVisitor. This provides sufficient additional context for the binary comparison ops (and "IN" membership checks) to enable automatic temporal string casts (which PostgreSQL syntax allows).

Handles comparison with bare column names (eg: col > '2020-02-20') and cast expressions (eg: expr::date > '2020-02-20'), which catches most typical usage. We should be able to enhance further as needed.

Also:

Added col::datetime cast support (in addition to existing col::timestamp).

Example

import polars as pl

df = pl.DataFrame({
    "idx": [0, 1, 2],
    "dtm": [
        datetime(2024, 1, 7, 1, 2, 3, 123456),
        datetime(2006, 1, 1, 23, 59, 59, 555555),
        datetime(2020, 12, 30, 10, 30, 45, 987654),
    ],
    "dt": [
        date(2020, 12, 30),
        date(2077, 1, 1),
        date(1960, 1, 7),
    ],
})

df.sql("SELECT * FROM self WHERE dtm < '2010-01-01 01:02:03'")
# shape: (1, 3)
# ┌─────┬────────────────────────────┬────────────┐
# │ idx ┆ dtm                        ┆ dt         │
# │ --- ┆ ---                        ┆ ---        │
# │ i64 ┆ datetime[μs]               ┆ date       │
# ╞═════╪════════════════════════════╪════════════╡
# │ 1   ┆ 2006-01-01 23:59:59.555555 ┆ 2077-01-01 │
# └─────┴────────────────────────────┴────────────┘

df.sql("SELECT * FROM self WHERE dt IN ('1960-01-07','2077-01-01','2222-02-22')")
# shape: (2, 3)
# ┌─────┬────────────────────────────┬────────────┐
# │ idx ┆ dtm                        ┆ dt         │
# │ --- ┆ ---                        ┆ ---        │
# │ i64 ┆ datetime[μs]               ┆ date       │
# ╞═════╪════════════════════════════╪════════════╡
# │ 1   ┆ 2006-01-01 23:59:59.555555 ┆ 2077-01-01 │
# │ 2   ┆ 2020-12-30 10:30:45.987654 ┆ 1960-01-07 │
# └─────┴────────────────────────────┴────────────┘

[alexander-beedie]

Need to check/fix a few tests in the morning; parking in Draft in the meantime…

Update: done.

[eitsupi]

I suppose this is not a specific to python? (feat: instead of feat(python):)

[alexander-beedie]

I suppose this is not a specific to python? (feat: instead of feat(python):)

Quite right; brain clearly on autopilot there - fixed ;)

[codspeed-hq]

CodSpeed Performance Report

Merging #15958 will not alter performance

_{Comparing alexander-beedie:sql-implicit-temporal-strings (86e7c7e) with main (c5489ac)}

Summary

✅ 35 untouched benchmarks

[codecov]

Codecov Report

Attention: Patch coverage is 94.11765% with 5 lines in your changes are missing coverage. Please review.

Project coverage is 80.92%. Comparing base (c5489ac) to head (7af4bfe).
Report is 2 commits behind head on main.

❗ Current head 7af4bfe differs from pull request most recent head 86e7c7e. Consider uploading reports for the commit 86e7c7e to get more accurate results

Files	Patch %	Lines
crates/polars-sql/src/sql_expr.rs	93.65%	4 Missing ⚠️
crates/polars-sql/src/functions.rs	90.90%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #15958      +/-   ##
==========================================
- Coverage   80.98%   80.92%   -0.06%     
==========================================
  Files        1386     1385       -1     
  Lines      178479   178277     -202     
  Branches     2877     3050     +173     
==========================================
- Hits       144539   144276     -263     
- Misses      33448    33514      +66     
+ Partials      492      487       -5     

Flag	Coverage Δ
python	74.41% <87.05%> (?)
rust	78.14% <94.11%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ritchie46]

Nice one @alexander-beedie. I've left some comments.