Treat explicit None filter as unrestricted for factor_library

Passing None for any factor_library filter column now removes that
filter entirely and returns all values for that column (e.g.
min_size_quantile=None includes all size groups), mirroring
purrr::compact() on NULL in the R package. Default None values (e.g.
n_portfolios_secondary) still match rows where the column is null.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: christophscheuch

CHANGELOG.md

@@ -7,7 +7,10 @@
   `sorting_variable` now returns the default portfolio construction for
   all sorting variables instead of raising an error. Other defaults
   (e.g. `rebalancing`, `weighting_scheme`) still apply; pass
-  `fill_all=True` to leave every column unrestricted.
+  `fill_all=True` to leave every column unrestricted. Passing `None` for
+  any filter column now removes that filter entirely and returns all
+  values for that column (e.g. `min_size_quantile=None` includes all
+  size groups), matching the R package's `NULL` behavior.
 - **Dependencies (replaced pyfixest with formulaic):** The `pyfixest` dependency was dropped in favor of [`formulaic`](https://github.com/matthewwardrop/formulaic) plus a small internal numpy OLS helper (`_fit_ols`). `pyfixest` was used only for plain OLS with classical (IID) standard errors in `estimate_model` and the cross-sectional / IID-variance steps of `estimate_fama_macbeth`, but it pulled in `great-tables` → `multimark`, a `cffi` C-extension that ships no Python 3.14 wheels and therefore required a C/C++ toolchain (e.g. MSVC Build Tools on Windows) to install on unsupported interpreters. `_fit_ols` builds the design matrix via `formulaic` and reproduces `feols` coefficients, standard errors, t-statistics, and residuals to ~1e-9 for models without fixed effects, so results are unchanged. Note that `estimate_model` and `estimate_fama_macbeth` now perform classical OLS only (the fixed-effects / clustered-SE features of `pyfixest` were never used and are no longer available).
 
 ## v0.1.0

tests/test_download_data_huggingface.py

@@ -284,6 +284,45 @@ def test_sorting_variable_optional_returns_all_with_defaults():
     assert ids == [1, 2]
 
 
+def test_explicit_none_removes_filter_returning_all_values():
+    """Test passing None for a column returns all values for it."""
+    grid = pd.DataFrame(
+        {
+            "id": [1, 2, 3],
+            "sorting_variable": ["sv_me", "sv_me", "sv_me"],
+            "min_size_quantile": [0.2, 0.4, 0.6],
+            "exclude_financials": [False, False, False],
+            "exclude_utilities": [False, False, False],
+            "exclude_negative_earnings": [False, False, False],
+            "sorting_variable_lag": ["6m", "6m", "6m"],
+            "rebalancing": ["monthly", "monthly", "monthly"],
+            "n_portfolios_main": [10, 10, 10],
+            "sorting_method": ["univariate", "univariate", "univariate"],
+            "n_portfolios_secondary": [None, None, None],
+            "breakpoints_exchanges": ["NYSE", "NYSE", "NYSE"],
+            "breakpoints_min_size_threshold": [None, None, None],
+            "weighting_scheme": ["VW", "VW", "VW"],
+        }
+    )
+    available = pd.DataFrame({"path": ["grid.parquet"], "size": [100]})
+    with (
+        patch(
+            "tidyfinance.download_tidy_finance._get_available_huggingface_files",
+            return_value=available,
+        ),
+        patch(
+            "tidyfinance.download_tidy_finance.pd.read_parquet",
+            return_value=grid,
+        ),
+    ):
+        ids = _filter_factor_library_grid(
+            sorting_variable="me", min_size_quantile=None
+        )
+
+    # The default 0.2 screen is removed, so all size groups are returned.
+    assert ids == [1, 2, 3]
+
+
 def test_fill_all_false_defaults_applied_row_filtered_out():
     """Test fill_all = FALSE: defaults applied, row filtered out."""
     grid = pd.DataFrame(

tidyfinance/download_tidy_finance.py

@@ -269,7 +269,10 @@ def _filter_factor_library_grid(fill_all: bool = False, **filters) -> list:
     **filters : dict
         Named arguments of the form 'column=value' used to filter the
         grid. Each value may be a scalar or a list/tuple to match multiple
-        levels. Supported columns and their defaults are:
+        levels. Passing 'None' for a column removes that filter entirely,
+        returning all values for that column (e.g.,
+        'min_size_quantile=None' includes all size groups). Supported
+        columns and their defaults are:
 
         - 'sorting_variable': no default. When omitted, all sorting
           variables are returned (subject to the remaining defaults).
@@ -311,8 +314,19 @@ def _filter_factor_library_grid(fill_all: bool = False, **filters) -> list:
                 "n_portfolios_secondary must be provided."
             )
 
+    # A filter the caller explicitly sets to None is removed entirely so
+    # all values for that column are returned, mirroring purrr::compact()
+    # on NULL in the R implementation. These columns are recorded so the
+    # defaults below do not reintroduce a filter for them. Default None
+    # values (e.g. n_portfolios_secondary) are applied afterwards and
+    # instead match rows where the column is null.
+    unrestricted = {col for col, value in filters.items() if value is None}
+    filters = {col: v for col, v in filters.items() if v is not None}
+
     if not fill_all:
-        filters = {**_FACTOR_LIBRARY_DEFAULTS, **filters}
+        for col, default in _FACTOR_LIBRARY_DEFAULTS.items():
+            if col not in filters and col not in unrestricted:
+                filters[col] = default
 
     grid = _download_factor_library_grid().assign(
         sorting_variable=lambda x: x["sorting_variable"].str.replace(
@@ -611,6 +625,10 @@ def _download_data_huggingface(
     - 'weighting_scheme' (defaults to 'VW'): return weighting within
       portfolios; 'VW' for value-weighted or 'EW' for equal-weighted.
 
+    Passing 'None' for any filter column removes that filter entirely,
+    returning all values for that column (e.g., 'min_size_quantile=None'
+    includes all size groups).
+
     Parameters
     ----------
     dataset : str