Repository navigation
Triage: 11 open issues appear resolved on master #75
Copy link
Copy link
Closed
Description
Verification completed (2026-09-29, published gtfparse 2.9.1): Ten stale reports were verified and closed; #19 remains open pending the original input. A current Polars dependency-floor bug is tracked separately in #85. See the full verification results. This triage tracker is complete; the original 2.8.0 notes below are historical.
Of the 21 open issues, 11 look fixed on master (2.8.0). I checked each one against the code and, where the report had input, ran it with polars 1.43 / pyarrow 25.
Fixed; can be closed
- read_gtf has ERROR "scan_csv() got an unexpected keyword argument 'sep'" #36, BackTrace for exception when parsing gtf file #38
scan_csv() got an unexpected keyword argument 'sep': fixed by New polars version & misc improvements #42 (see the maintainer's comment on read_gtf has ERROR "scan_csv() got an unexpected keyword argument 'sep'" #36).read_gtf.py:118now passesseparator. - Please do not use logging basicConfig for library software #40
logging.basicConfigat import: removed in 9e9222d (Fix #70: stop calling logging.basicConfig() at import time #71). There is nobasicConfiganywhere ingtfparse/. - DeprecationWarning: The argument
dtypesforread_csvis deprecated. #50dtypesdeprecation:read_gtf.py:121passesschema_overrides(5a740c9). - Error in gtfparse when calling vcf-expression-annotator #49
'LazyFrame' object has no attribute 'lazy'(reported on 2.0.1 / Python 3.6):read_gtf.py:124now uses eagerread_csv(...).lazy(), andrequires-python >= 3.9. - gtfparse.parsing_error.ParsingError: Integer column has NA values in column 3 #19 Scallop GTF, "Integer column has NA values": the snippet from the issue parses into 30 rows × 13 columns. That error message no longer exists in the code.
- Conversion to Pandas in read_gtf has missing dependency #32 pyarrow undeclared, pandas is imported but isn't listed in https://github.com/openvax/gtfparse #52 pandas undeclared: both are in
requirements.txt:2-3. - Update gtfparse to support pyarrow >=0.15 #47 pyarrow
<14.1pin: upper bounds removed in 1c50944 (Update dependencies #51). - ImportError: cannot import name table_to_blockmanager #45
table_to_blockmanager(pyarrow 10) and Pinnumpyto 1.X.X #55 numpy 2_ARRAY_API(old pyarrow): both come from old pyarrow builds, andpyarrow>=18.0.0is now required. The commenter on ImportError: cannot import name table_to_blockmanager #45 says upgrading pyarrow fixed it.
Not in this repo
- Bioconda wrong dependencies for version 2.1.0 #43: Bioconda recipe for 2.1.0.
Still reproduce
- Fix bug for records with ";" in the value #44:
description "part1; part2"parses aspart1. - Attribute parsing fix #23:
attribute_parsing.py:71still doessplit(" ", 2)[:2].
Not checked: #16, #53, #65, #66, #69. #13 is addressed by #74.
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels
Activity
iskandr commentedon Sep 29, 2026
Verified against the published gtfparse 2.9.1 package on 2026-09-29. Ten of the eleven suspected-resolved reports are now closed, each with its own verification note. #19 remains open because the original input has not been reproduced faithfully.
read_csv(separator=...); Ensembl parsing succeeds without the oldseperror..lazy()on one. Current metadata requires Python >=3.9.schema_overridesreplacesdtypes; parsing and pandas conversion pass with deprecations as errors.logging.basicConfigand preserves both unconfigured and application-configured root logging.pyarrow>=18.0.0andpandas>=2.1.0; imports and conversion pass.pyarrow>=18.0.0. Parsing/conversion succeeds with 25.0.1 as well as 18.0.0.table_to_blockmanager/_ARRAY_APIfailures do not reproduce on these supported combinations.Wrong number of columns. Normalizing only field separators produces 30 rows × 13 columns, but the originalhybrid.gtffile is unavailable. That normalized check does not establish that the original report is resolved.Checks ran in an isolated Python 3.12 environment installed from PyPI, with
pip checkpassing andDeprecationWarningtreated as an error. The tested dependency combinations were:Fixture checks cover Ensembl (995 rows), StringTie (1,000 rows), RefSeq (24 rows), and the separator-normalized Scallop example (30 rows). Each was read into Polars and/or pandas as appropriate; the logging and lazy-reader checks exercise the installed release as well.
New current bug found: #85. The declared
polars>=0.20.2floor is too low for the unconditionalschema_overridescall. Polars 0.20.2 installs withpip checkpassing but fails on a valid one-line GTF. Polars 0.20.31 passes the targeted audit checks; its full API support still needs verification before changing the lower bound. The old Polars report closures link this separate current issue.Other dispositions from the original tracker: #23 and #44 shipped in 2.9.0 via #82; #53 shipped in 2.9.1 via #84. #43 remains a separate Bioconda packaging report. #13/#74, #16, #37, and feature/design work #65/#66/#69 remain separate open work and were not closed by this audit.
Closing this triage tracker with the individual dispositions recorded. Follow-up is #85, plus the original input needed for #19.