feat(cli): add metaxy metadata read command

[AnkitSharma-29]

This PR implements the metadata read command as requested in #706. It allows users to quickly execute arbitrary SQL queries against feature metadata using DuckDB and Polars. The command supports a simplified interface with --filter and --select arguments, as well as an advanced --query argument. Output can be directed to stdout or a file in markdown, json, csv, or parquet formats. Comprehensive unit tests have been added in tests/cli/test_cli_metadata_read.py to ensure all functionality works as expected.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!

[danielgafni]

Thanks!

Overall looking good, I suggested a few improvements.

Also, please improve the tests to assert more about the data - ideally the full dataframe for Parquet (json and csv will probably lose proper datatypes)

[danielgafni]

Left a bunch of comments

Overall this looks suspiciously AI-ish, please review the output better

[danielgafni]

1. Minor comments
2. Please consolidate Literal["csv", "json", "parquet", "markdown"] into WriteFormat: TypeAlias = ... at the top of this module
3. I think the tests should be refactored:
   1. pamaterization: instead of defining a single test for every combinations of format and output parameters, make a single test using @pytest.mark.parametrize for both parameters.
   2. data: I think it should be possible to make a single shared @pytest.fixture for the datafarme used in these tests.
4. When --query is used with a non-Ibis store, please let's not use Polars at all. Just use duckdb instead. We can fail with a clear message if duckdb is not available. This is because Polars' support for SQL is way behind DuckDB's.
5. Since 0.1.7 has been released and Metaxy aims to eventually support Map types by default, let's replace all collect_to_polars with _collect_to_arrow, and use DuckDB instead of Polars for writes. For Markdown we can use the Markdown extension. Avoid using Polars completely, let's do not have the import polars at all. This is because Polars doesn't support the Map type.

I am documenting references to all of these docs for educational purposes since previously you said you wanted to learn. I assume good faith here so please take your time to go though them and understand what's expected without AI assistance :) As you can see AI is struggling to write clean (or even correct at all) code without proper human guidance and is useless without it.

[AnkitSharma-29]

1. Minor comments

2. Please consolidate Literal["csv", "json", "parquet", "markdown"] into WriteFormat: TypeAlias = ... at the top of this module

3. I think the tests should be refactored:

   1. pamaterization: instead of defining a single test for every combinations of format and output parameters, make a single test using @pytest.mark.parametrize for both parameters.
   2. data: I think it should be possible to make a single shared @pytest.fixture for the datafarme used in these tests.

4. When --query is used with a non-Ibis store, please let's not use Polars at all. Just use duckdb instead. We can fail with a clear message if duckdb is not available. This is because Polars' support for SQL is way behind DuckDB's.

5. Since 0.1.7 has been released and Metaxy aims to eventually support Map types by default, let's replace all collect_to_polars with _collect_to_arrow, and use DuckDB instead of Polars for writes. For Markdown we can use the Markdown extension. Avoid using Polars completely, let's do not have the import polars at all. This is because Polars doesn't support the Map type.

I am documenting references to all of these docs for educational purposes since previously you said you wanted to learn. I assume good faith here so please take your time to go though them and understand what's expected without AI assistance :) As you can see AI is struggling to write clean (or even correct at all) code without proper human guidance and is useless without it.

Thanks for the docs really appreciate it. I’ll go through them and make sure I understand everything before updating the PR.

I’m trying to keep AI usage minimal. Earlier I got stuck while updating the PR and used Antigravity, which may have unintentionally changed some code. I’ll review and fix it properly.

[danielgafni]

@AnkitSharma-29 fyi once #1130 lands you'll need to use it as from metaxy._utils import collect_to_arrow

[danielgafni]

Thanks, two small comments left!

[danielgafni]

I think this should just fail with an error. The current fallback doesn't produce Markdown while the user has clearly requested it.

[AnkitSharma-29]

Good catch....removed the fallback entirely so it no longer emits non-markdown output. While fixing it I found the install_extension("markdown") call was the actual cause of the fallback being hit, and it's unnecessary since markdown is rendered via pandas to_markdown(). It now produces real Markdown and fails loudly if rendering can't happen. Added a test covering the failure case.

[danielgafni]

Please do not use Pandas. Just fix the markdown dudkcb extension installation instead (if that's a thing). Avoid Pandas at all cost.

[AnkitSharma-29]

Removed pandas from the markdown output it's now generated natively from the DuckDB Arrow result, no to_markdown(). There's no installable DuckDB markdown extension, so the table is built inline. pushed in ddff9f4.

[AnkitSharma-29]

Thanks, two small comments left!

Thanks! Both addressed in 261bec5.