fix: check_code misses lines_before_imports-only changes (#2242)

[binggao1230]

Problem

isort.check_code (and isort.core.process) returns True (no changes needed) when lines_before_imports requires adding or removing blank lines before the import block, but the imports themselves are already correctly sorted.

Closes #2242.

import isort

code = "from a import b, x\n\nfoo = 'bar'\n"

# isort.code() correctly adds the blank line:
assert isort.code(code, lines_before_imports=1) == "\n" + code  # ✓

# But check_code incorrectly says no change is needed:
assert isort.check_code(code, lines_before_imports=1)  # ✗ should be False

Root cause

Two cooperating issues in isort/core.py:

1. _has_changed() used .strip() for comparison — this stripped leading newlines that output.sorted_imports had prepended (via lines_before_imports), making additions invisible:

# Before: both sides stripped → leading blank line difference hidden
return before.strip() != after.strip()

2. Blank lines consumed before the import block were excluded from the comparison — when lines_before_imports >= 0, blank lines preceding the import block are buffered in lines_before and then discarded (not written to the output stream). The raw_import_section passed to _has_changed never included those consumed lines, so removing them was also never detected as a change.

Fix

• Change _has_changed to use .rstrip() instead of .strip(), preserving leading whitespace differences while still ignoring insignificant trailing newlines.
• Capture the buffered pre-import blank lines in above_import_section and prepend them to raw_import_section before the comparison, so removed blank lines are detected.

Tests

Added test_check_code_detects_lines_before_import_changes covering:

• Adding a blank line (0 present → 1 required)
• Removing a blank line (1 present → 0 required)
• Already-correct cases (no false positives)

Note: PR #2541 addresses the same issue with only the .rstrip() fix, which handles the "adding blank lines" case. This PR also fixes the "removing blank lines" case via the above_import_section addition, and includes the requested test.

This pull request was prepared with the assistance of AI, under my direction and review.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.33%. Comparing base (0740d0c) to head (8601df1).
⚠️ Report is 50 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2563      +/-   ##
==========================================
+ Coverage   99.16%   99.33%   +0.17%     
==========================================
  Files          41       41              
  Lines        3101     3157      +56     
  Branches      671      682      +11     
==========================================
+ Hits         3075     3136      +61     
  Misses         14       14              
+ Partials       12        7       -5     

🚀 New features to boost your workflow:

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

[DanielNoord]

Did you see #2541? And if so, why should we continue with this approach instead of that one?