Phase 1: eliminate redundant data copies and add optional pyarrow engine#26
Open
wsnoble wants to merge 4 commits into
Open
Phase 1: eliminate redundant data copies and add optional pyarrow engine#26wsnoble wants to merge 4 commits into
wsnoble wants to merge 4 commits into
Conversation
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
wsnoble
force-pushed
the
scale-phase-1-2
branch
3 times, most recently
from
92a5f1f to
602d9b6
Compare
There was a problem hiding this comment.
Pull request overview
This PR reduces memory overhead when constructing/inspecting PsmDataset and speeds up text parsing by optionally using pandas’ pyarrow CSV engine (with fallback).
Changes:
- Remove deep-copy behavior from
PsmDataset.dataand changecopy_datadefaults toFalseacross dataset/parsers. - Add optional
engine="pyarrow"usage forread_csvin text parsing helpers.
Reviewed changes
Copilot reviewed 8 out of 8 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| crema/utils.py | Adds optional pyarrow-based read_csv path in parse_psms_txt() |
| crema/parsers/txt.py | Adds optional pyarrow-based read_csv path in _parse_psms() and changes read_txt(copy_data) default |
| crema/parsers/tide.py | Changes read_tide(copy_data) default to False |
| crema/parsers/msgf.py | Changes read_msgf(copy_data) default to False |
| crema/parsers/msfragger.py | Changes read_msfragger(copy_data) default to False |
| crema/parsers/msamanda.py | Changes read_msamanda(copy_data) default to False |
| crema/parsers/comet.py | Changes read_comet(copy_data) default to False |
| crema/dataset.py | Changes copy_data default to False and stops copying on .data access |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+88
to
+94
| kwargs = dict( | ||
| sep="\t", skiprows=int(skip_line), usecols=lambda c: c in cols | ||
| ) | ||
| try: | ||
| return pd.read_csv(txt_file, engine="pyarrow", **kwargs) | ||
| except ImportError: | ||
| return pd.read_csv(txt_file, **kwargs) |
Comment on lines
46
to
+50
| If true, a deep copy of the data is created. This uses more memory, but | ||
| is safer because it prevents accidental modification of the underlying | ||
| data. This argument only has an effect when `pin_files` is a | ||
| :py:class:`pandas.DataFrame` | ||
| :py:class:`pandas.DataFrame`. Default is ``False`` to minimise memory | ||
| use; set to ``True`` if you need to protect the original DataFrame. |
Comment on lines
118
to
+120
| def data(self): | ||
| """The collection of PSMs as a :py:class:`pandas.DataFrame`.""" | ||
| return self._data.copy() | ||
| return self._data |
| # pyarrow engine doesn't support callable usecols, so pass an explicit list. | ||
| # Fall back to the default C engine if pyarrow is missing or the pandas | ||
| # version doesn't support engine="pyarrow" (raises ImportError or ValueError). | ||
| with open(txt_file) as fh: |
Comment on lines
+101
to
+103
| fh.readline() | ||
| header = fh.readline().rstrip("\n").split("\t") | ||
| explicit_cols = [c for c in header if c in cols] |
Comment on lines
+119
to
+124
| """The collection of PSMs as a :py:class:`pandas.DataFrame`. | ||
|
|
||
| Returns the internal DataFrame by reference. Callers that need an | ||
| independent copy should call ``.data.copy()`` explicitly. | ||
| """ | ||
| return self._data |
- Remove `.copy()` from `PsmDataset.data` property; callers that need an independent copy must now call `.data.copy()` explicitly. - Change `copy_data` default from `True` to `False` in `PsmDataset.__init__` and in all public parser entry-points (`read_txt`, `read_tide`, `read_comet`, `read_msamanda`, `read_msfragger`, `read_msgf`) so that file-backed DataFrames are never copied a second time on ingestion. - Use pyarrow CSV engine in `_parse_psms()` and `parse_psms_txt()` when pyarrow is available, falling back to the default C engine otherwise. This cuts parse time roughly 2–4× on large files at no API cost. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- utils.py: pyarrow engine doesn't support callable usecols; pre-compute an
explicit column list from the file header and catch (ImportError, ValueError)
so older pandas versions fall back cleanly to the C engine.
- parsers/txt.py: also catch ValueError (not just ImportError) in pyarrow
fallback, for consistency with utils.py and older pandas support.
- dataset.py: fix copy_data docstring ('pin_files' → 'psms'); document that
the data property returns the internal DataFrame by reference.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Open the file with encoding='utf-8' (matching pandas read_csv default) when pre-reading the header for the pyarrow usecols list. If a UnicodeDecodeError occurs, fall back directly to the C engine with a callable usecols filter, avoiding a crash on non-UTF-8 locale files. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- confidence.py: self._data = psms.data.copy() to avoid mutating the
original PsmDataset when confidence assigns columns (e.g. "pairing")
- utils.py: rstrip("\r\n") instead of "\n" so Windows CRLF files are
parsed correctly by the pyarrow engine path
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
.copy()fromPsmDataset.dataproperty — previously allocated a full copy of the dataset on every property access.copy_datadefault toFalseinPsmDataset.__init__and all parser entry-points (read_txt,read_tide,read_comet,read_msamanda,read_msfragger,read_msgf) — eliminates a redundant deep copy when callers pass an existing DataFrame._parse_psms()andparse_psms_txt()— 2–4× faster parsing when pyarrow is installed.All 90 unit tests pass. This is Phase 1 of
SCALING_PLAN.md.Test plan
read_tide/read_txtstill work on a real Tide file🤖 Generated with Claude Code