polars-config-meta

A Polars plugin for persistent DataFrame-level metadata.

polars-config-meta offers a simple way to store and propagate Python-side metadata for Polars DataFrames. It achieves this by:

• Registering a custom config_meta namespace on each DataFrame (via @register_dataframe_namespace).
• Keeping an internal dictionary keyed by the id(df), with automatic weak-reference cleanup to avoid memory leaks.
• Providing a “fallthrough” mechanism so you can write df.config_meta.some_polars_method(...) and have the resulting new DataFrame automatically inherit the old metadata—no manual copying required.
• Optionally embedding that metadata in file‐level Parquet metadata when you call df.config_meta.write_parquet(...), and retrieving it with read_parquet_with_meta(...).

Installation

pip install polars-schema-index[polars]

On older CPUs add the polars-lts-cpu extra:

pip install polars-schema-index[polars-lts-cpu]

For parquet file-level metadata read/writing, add the pyarrow extra:

pip install polars-schema-index[pyarrow]

Key Points

1. No Monkey-Patching or Subclassing We do not modify Polars’ built-in classes at runtime or create a custom subclass of DataFrame. Everything is implemented through a plugin namespace.

2. Weak-Reference Based We store metadata in class-level dictionaries keyed by id(df) and hold a weakref to the DataFrame. Once the DataFrame is garbage-collected, the metadata is removed too.

3. Automatic Metadata Copying

   • When you call df.config_meta.with_columns(...) (or any other Polars method) through the config_meta namespace, we intercept the result.
   • If it’s a new DataFrame, the plugin copies the old one’s metadata forward.

4. Parquet Integration

   • df.config_meta.write_parquet("file.parquet") automatically embeds the plugin metadata into the Arrow schema’s metadata.
   • read_parquet_with_meta("file.parquet") reads the file, extracts that metadata, and reattaches it to the returned DataFrame.

5. Opt-In Only

   • If you call df.with_columns(...) without .config_meta. in front, Polars has no knowledge of this plugin, so metadata will not copy forward.
   • If you want transformations to preserve metadata, call them via df.config_meta.<method>(...).

Basic Usage

import polars as pl
import polars_config_meta  # this registers the plugin

df = pl.DataFrame({"a": [1, 2, 3]})
df.config_meta.set(owner="Alice", confidence=0.95)

# Use the plugin to transform; the returned DataFrame inherits metadata:
df2 = df.config_meta.with_columns(doubled=pl.col("a") * 2)
print(df2.config_meta.get_metadata())
# -> {'owner': 'Alice', 'confidence': 0.95}

# Write to Parquet, storing the metadata in file-level metadata:
df2.config_meta.write_parquet("output.parquet")

# Later, read it back:
from polars_config_meta import read_parquet_with_meta
df_in = read_parquet_with_meta("output.parquet")
print(df_in.config_meta.get_metadata())
# -> {'owner': 'Alice', 'confidence': 0.95}

Storage and Garbage Collection

Internally, the plugin stores metadata in a global dictionary, _df_id_to_meta, keyed by id(df), and also keeps a weakref to each DataFrame. As soon as a DataFrame is out of scope and garbage-collected, the entry in _df_id_to_meta is automatically removed. This prevents memory leaks and keeps the plugin usage simple.

Common Patterns

• Setting Metadata: df.config_meta.set(key1="val1", key2="val2", ...)

• Retrieving Metadata: df.config_meta.get_metadata() (returns a dict)

• Updating Metadata From a Dict: df.config_meta.update({"some_key": "new_val", ...})

• Merging Metadata From Other DataFrames:

  df3 = pl.DataFrame(...)
  df3.config_meta.merge(df1, df2)

  This copies all key–value pairs from df1 and df2 into df3’s metadata.

• Transformations

  • df.config_meta.with_columns(...)
  • df.config_meta.select(...)
  • df.config_meta.filter(...)
  • etc.

For any method that returns a new DataFrame, the plugin copies metadata forward. If the method returns something else (like a Series or plain Python object), the plugin does nothing.

Caveats

• Must Use df.config_meta.<method> If you call Polars methods directly on df, the plugin can’t intercept the result, so metadata will not be inherited.
• Not Official Polars Feature This is purely at the Python layer. Polars doesn’t guarantee stable IDs or official hooks for such metadata.
• Arrow/IPC/CSV For other formats, you’d need to write your own logic to embed or retrieve the metadata. Currently, only Parquet is supported out of the box via df.config_meta.write_parquet and read_parquet_with_meta.

Contributing

1. Issues & Discussions: Please open a GitHub issue for bugs, ideas, or questions.
2. Pull Requests: PRs are welcome! This plugin is a community-driven approach to persist DataFrame-level metadata in Polars.

License

This project is licensed under the MIT License.