bf: validate --eval-filter up-front, fail with a clear error (#440)

[yarikoptic]

• Fixes running con-duct ls on Mac leads to many error messages #440.

Summary

con-duct ls -e EXPR used to call eval(EXPR, ...) inside the per-file loop
of load_duct_runs(), inside a broad except Exception block. The reporter
in #440 had a hidden U+00A0 non-breaking space in their filter expression
(easy to produce on macOS with Option+Space, or when copy-pasting from a
browser/document). The resulting SyntaxError was caught, misattributed
to each JSON file, and logged as:

Failed to load file <_io.TextIOWrapper name='.duct/logs/…_info.json'>: invalid non-printable character U+00A0 (<string>, line 1)

— which sent the user hunting through their JSON logs for corruption that
was never there.

This PR centralises filter compilation so a bad expression fails once,
up-front, with a clear message, before any file is opened.

• New compile_eval_filter() helper compiles the expression once via
  compile(..., '<eval-filter>', 'eval'). On SyntaxError it raises
  ValueError naming the offending expression and explicitly hinting at
  U+00A0 / macOS Option+Space as a common cause.
• ls() calls it first thing, catches ValueError, logs it via
  lgr.error, and returns exit code 2 — no traceback, no per-file noise.
• load_duct_runs() now accepts either a string (which it compiles) or a
  pre-compiled CodeType, and reuses it in the per-file eval() loop —
  also avoids re-parsing the expression once per file.

After the fix, the reporter's command produces:

[ERROR   ] con_duct.ls: Invalid --eval-filter expression '"fmriprep" in command and\xa0exit_code == 1': invalid non-printable character U+00A0 (<eval-filter>, line 1). Check for hidden non-printable characters (e.g. U+00A0 non-breaking space produced by macOS Option+Space or copy-paste from a browser/document).

and exits with code 2.

Test plan

• tox -e py3 -- test/test_ls.py — 29 passed (23 pre-existing + 6 new)
• Full suite: tox -- test/ — 397 passed
• tox -e lint — clean
• tox -e typing — clean
• Manual repro of the running con-duct ls on Mac leads to many error messages #440 command (with a real U+00A0 injected via
  printf '%b') now prints the single ERROR above and exits 2.

New tests added to test/test_ls.py:

• test_compile_eval_filter_none_returns_none
• test_compile_eval_filter_valid
• test_compile_eval_filter_rejects_nbsp (regression test for running con-duct ls on Mac leads to many error messages #440)
• test_compile_eval_filter_rejects_syntax_error
• test_load_duct_runs_fails_fast_on_bad_filter — verifies no file is
  even opened when the filter is invalid, and no per-file "Failed to load
  file" warning is emitted.
• test_ls_exits_cleanly_on_bad_filter — verifies exit code 2 and a
  user-facing ERROR log line mentioning U+00A0.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.01%. Comparing base (cc80272) to head (5f1d59b).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #441      +/-   ##
==========================================
- Coverage   91.08%   91.01%   -0.07%     
==========================================
  Files          15       15              
  Lines        1245     1258      +13     
  Branches      170      171       +1     
==========================================
+ Hits         1134     1145      +11     
- Misses         77       78       +1     
- Partials       34       35       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR fixes confusing per-file “Failed to load file …” noise when con-duct ls --eval-filter contains a syntax error (notably hidden U+00A0 non-breaking spaces), by compiling/validating the expression once up-front and failing with a single clear error message and exit code 2.

Changes:

• Add compile_eval_filter() to compile --eval-filter once (with <eval-filter> filename) and convert SyntaxError into a user-facing ValueError.
• Update ls() to validate the filter before globbing/opening any files and to exit cleanly (rc=2) on invalid expressions.
• Update load_duct_runs() to accept a precompiled CodeType and reuse it in the per-file eval() loop; add tests covering the new behavior and the gh-440 regression.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
src/con_duct/ls.py	Adds upfront filter compilation and updates ls() / load_duct_runs() to reuse a compiled filter and fail cleanly on invalid expressions.
test/test_ls.py	Adds unit/regression tests for filter compilation, fast failure before file I/O, and ls() exit code/logging behavior.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.