From 929e83d4069453afd7cf40d6a124719a31f29fed Mon Sep 17 00:00:00 2001
From: Alex Rubinsteyn <alex.rubinsteyn@gmail.com>
Date: Wed, 13 May 2026 14:40:42 -0400
Subject: [PATCH] Fix #56: NameError on expand_attribute_column=False
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The else branch in read_gtf passed `result_df` to parse_gtf before
`result_df` had been assigned — a guaranteed NameError that survived
because the branch was never tested. Replaced with the correct
`filepath_or_buffer` argument and also pass `split_attributes=False`
so the unexpanded path doesn't emit the helper `attribute_split` column
the user didn't ask for.

Added tests/test_expand_attribute_column_false.py covering pandas /
polars / dict result types, the features filter, and the alias /
version-cast kwargs as graceful no-ops when attribute columns aren't
expanded.

Bumped to 2.7.1.
---
 gtfparse/__init__.py                        |  2 +-
 gtfparse/read_gtf.py                        |  5 +-
 tests/test_expand_attribute_column_false.py | 74 +++++++++++++++++++++
 3 files changed, 79 insertions(+), 2 deletions(-)
 create mode 100644 tests/test_expand_attribute_column_false.py

diff --git a/gtfparse/__init__.py b/gtfparse/__init__.py
index bb8908b..436a0de 100644
--- a/gtfparse/__init__.py
+++ b/gtfparse/__init__.py
@@ -23,7 +23,7 @@
     read_gtf,
 )
 
-__version__ = "2.7.0"
+__version__ = "2.7.1"
 
 __all__ = [
     "GENCODE_BIOTYPE_ALIASES",
diff --git a/gtfparse/read_gtf.py b/gtfparse/read_gtf.py
index 23e2647..13ddc84 100644
--- a/gtfparse/read_gtf.py
+++ b/gtfparse/read_gtf.py
@@ -354,7 +354,10 @@ def read_gtf(
             features=features,
         )
     else:
-        result_df = parse_gtf(result_df, features=features)
+        # When the caller opts out of attribute expansion they want the raw
+        # 'attribute' column verbatim — no need to also produce the
+        # 'attribute_split' helper that parse_gtf adds by default.
+        result_df = parse_gtf(filepath_or_buffer, features=features, split_attributes=False)
 
     # converting back to pandas here because Polars bugs manifest
     # as `pyo3_runtime.PanicException: assertion `left == right` failed: impl error`
diff --git a/tests/test_expand_attribute_column_false.py b/tests/test_expand_attribute_column_false.py
new file mode 100644
index 0000000..b06a206
--- /dev/null
+++ b/tests/test_expand_attribute_column_false.py
@@ -0,0 +1,74 @@
+"""Regression tests for #56: read_gtf(expand_attribute_column=False)
+used to raise NameError because the else branch referenced `result_df`
+before it had been assigned.
+"""
+
+import pandas as pd
+
+from gtfparse import read_gtf
+
+from .data import data_path
+
+GTF_PATH = data_path("ensembl_grch37.head.gtf")
+
+
+def test_expand_attribute_column_false_returns_raw_attribute_pandas():
+    df = read_gtf(GTF_PATH, expand_attribute_column=False, result_type="pandas")
+    assert isinstance(df, pd.DataFrame)
+    # raw attribute column is preserved verbatim
+    assert "attribute" in df.columns
+    # none of the per-key attribute columns are produced
+    assert "gene_name" not in df.columns
+    assert "transcript_id" not in df.columns
+    # the helper 'attribute_split' column from parse_gtf is also suppressed
+    assert "attribute_split" not in df.columns
+    # core GTF columns are present and populated
+    for col in ("seqname", "source", "feature", "start", "end", "strand"):
+        assert col in df.columns
+    assert len(df) > 0
+    # spot-check that the raw attribute string carries the original key/value form
+    assert any("gene_id" in val for val in df["attribute"].astype(str))
+
+
+def test_expand_attribute_column_false_returns_polars():
+    df = read_gtf(GTF_PATH, expand_attribute_column=False, result_type="polars")
+    # polars dataframe — has columns attribute but no per-key columns
+    assert "attribute" in df.columns
+    assert "gene_name" not in df.columns
+    assert "attribute_split" not in df.columns
+
+
+def test_expand_attribute_column_false_returns_dict():
+    result = read_gtf(GTF_PATH, expand_attribute_column=False, result_type="dict")
+    assert isinstance(result, dict)
+    assert "attribute" in result
+    assert "gene_name" not in result
+
+
+def test_expand_attribute_column_false_with_features_filter():
+    """The features filter must still apply when not expanding."""
+    df = read_gtf(
+        GTF_PATH,
+        expand_attribute_column=False,
+        features={"gene"},
+        result_type="pandas",
+    )
+    assert set(df["feature"]) == {"gene"}
+
+
+def test_expand_attribute_column_false_skips_alias_and_version_logic():
+    """When attribute columns aren't expanded, attribute_aliases has
+    nothing to rename and cast_version_columns has nothing to cast.
+    Neither should raise — both must be graceful no-ops on the raw
+    'attribute'-column-only frame."""
+    df = read_gtf(
+        GTF_PATH,
+        expand_attribute_column=False,
+        attribute_aliases={"gene_type": "gene_biotype"},
+        cast_version_columns=True,
+        result_type="pandas",
+    )
+    # alias source wasn't in columns → no rename happened → no canonical added
+    assert "gene_biotype" not in df.columns
+    # version columns weren't present → no cast → still nothing
+    assert "gene_version" not in df.columns