Avoid re-using imports beyond current edit site

[charliermarsh]

Given code like:

def a():
    try:
        pass
    except OSError:
        pass


import contextlib

Our current edit actions will attempt to form an edit to (1) use contextlib.suppress instead of try, and (2) retain the import contextlib at the bottom of the file. This leads to an out-of-order edit.

One fix here would be to simply reorder the edits. That would often be valid, but it could lead to confusing bugs if the function is called between the definition site and the import site. Another would be to add an import to the top of the file instead of re-using. This would also often be valid, but runs the risk that the import is shadowed between the top-of-file and any call sites. We don't currently have access to that kind of tracing. So, for now, I've opted to just avoid applying a fix in these cases.

Closes #4374.

[charliermarsh]

Quite torn on this. Maybe using the import is actually a better choice (and just sorting the edits).

[charliermarsh]

Hmm, actually, in the reported case, re-using the import would have led to broken code.

[charliermarsh]

I think the "right" fix is probably to add an import at top-of-file, if we can ensure that it's not shadowed later on.

[github-actions]

PR Check Results

Ecosystem

✅ ecosystem check detected no changes.

Benchmark

Linux

group                                      main                                   pr
-----                                      ----                                   --
linter/all-rules/large/dataset.py          1.06     14.7±0.09ms     2.8 MB/sec    1.00     13.8±0.13ms     2.9 MB/sec
linter/all-rules/numpy/ctypeslib.py        1.03      3.5±0.03ms     4.8 MB/sec    1.00      3.4±0.03ms     4.9 MB/sec
linter/all-rules/numpy/globals.py          1.02    418.8±2.53µs     7.0 MB/sec    1.00    412.0±1.83µs     7.2 MB/sec
linter/all-rules/pydantic/types.py         1.04      6.1±0.07ms     4.2 MB/sec    1.00      5.8±0.07ms     4.4 MB/sec
linter/default-rules/large/dataset.py      1.14      7.7±0.02ms     5.3 MB/sec    1.00      6.7±0.03ms     6.1 MB/sec
linter/default-rules/numpy/ctypeslib.py    1.10   1583.7±1.87µs    10.5 MB/sec    1.00   1441.7±6.07µs    11.5 MB/sec
linter/default-rules/numpy/globals.py      1.08    172.1±0.49µs    17.1 MB/sec    1.00    159.8±1.06µs    18.5 MB/sec
linter/default-rules/pydantic/types.py     1.11      3.4±0.01ms     7.6 MB/sec    1.00      3.0±0.01ms     8.4 MB/sec
parser/large/dataset.py                    1.01      5.4±0.03ms     7.5 MB/sec    1.00      5.3±0.00ms     7.6 MB/sec
parser/numpy/ctypeslib.py                  1.01   1045.6±0.73µs    15.9 MB/sec    1.00   1038.9±2.27µs    16.0 MB/sec
parser/numpy/globals.py                    1.01    106.1±0.19µs    27.8 MB/sec    1.00    105.1±0.22µs    28.1 MB/sec
parser/pydantic/types.py                   1.01      2.3±0.00ms    11.2 MB/sec    1.00      2.3±0.00ms    11.3 MB/sec

Windows

group                                      main                                   pr
-----                                      ----                                   --
linter/all-rules/large/dataset.py          1.03     22.4±0.90ms  1862.6 KB/sec    1.00     21.7±0.85ms  1917.6 KB/sec
linter/all-rules/numpy/ctypeslib.py        1.00      5.6±0.24ms     2.9 MB/sec    1.00      5.6±0.29ms     3.0 MB/sec
linter/all-rules/numpy/globals.py          1.08   717.1±88.39µs     4.1 MB/sec    1.00   663.0±84.64µs     4.5 MB/sec
linter/all-rules/pydantic/types.py         1.03      9.4±0.39ms     2.7 MB/sec    1.00      9.2±0.42ms     2.8 MB/sec
linter/default-rules/large/dataset.py      1.01     11.5±0.46ms     3.6 MB/sec    1.00     11.4±0.43ms     3.6 MB/sec
linter/default-rules/numpy/ctypeslib.py    1.00      2.4±0.15ms     7.0 MB/sec    1.00      2.4±0.11ms     7.0 MB/sec
linter/default-rules/numpy/globals.py      1.00   273.4±15.87µs    10.8 MB/sec    1.00   272.1±18.35µs    10.8 MB/sec
linter/default-rules/pydantic/types.py     1.00      5.1±0.28ms     5.0 MB/sec    1.00      5.1±0.23ms     5.0 MB/sec
parser/large/dataset.py                    1.01      9.0±0.34ms     4.5 MB/sec    1.00      8.9±0.36ms     4.5 MB/sec
parser/numpy/ctypeslib.py                  1.04  1761.0±84.21µs     9.5 MB/sec    1.00  1697.6±91.47µs     9.8 MB/sec
parser/numpy/globals.py                    1.00   178.1±10.22µs    16.6 MB/sec    1.00   178.2±14.75µs    16.6 MB/sec
parser/pydantic/types.py                   1.01      3.8±0.16ms     6.7 MB/sec    1.00      3.8±0.16ms     6.7 MB/sec