document performance characteristics

This commit is contained in:
2026-08-27 10:31:29 +02:00
parent 62a4fce866
commit 591bd86235

View File

@@ -51,8 +51,9 @@ would depend on configuration file ordering.
**Ties keep their original file order.** The sort keys only on the count, and
Python's sort is stable.
**Unchanged lines are still printed.** Dropping non-matching lines would
silently lose content, so they print last with a count of zero.
**Unchanged lines are still printed.** Dropping lines that matched nothing
would silently lose content, so every line is printed; those with no
replacements sort last.
**Only the line terminator is stripped.** `\n` and `\r\n` both; every other
character stays replaceable. The format has no quoting, so the trailing space in
@@ -61,22 +62,31 @@ character stays replaceable. The format has no quoting, so the trailing space in
## Performance characteristics
`replace_line` tries every pair at each position, so cost grows with both the
line length and the number of pairs. On a fixed 20,000-line input of
40-character lines, against two alternatives producing identical output:
line length and the number of pairs. Two alternatives, measured on the same
20,000-line input of 40-character lines and producing identical output:
| Pairs | this implementation | `re` alternation | prefix tree |
| ----: | ------------------: | ---------------: | ----------: |
| 3 | 0.27s | 0.011s | 0.067s |
| 100 | 6.13s | 0.011s | 0.073s |
| 1000 | 58.80s | 0.011s | 0.067s |
| Pairs | this implementation | regex alternation | prefix tree |
| ----: | ------------------: | ----------------: | ----------: |
| 3 | 1x | 1x | 1x |
| 100 | 20x | 1.2x | 1.1x |
| 1000 | 207x | 1.1x | 1.1x |
The regex is a constant-factor win, running the scan in C; the prefix tree is
the algorithmic one, taking the pair count out of the complexity entirely. The
explicit loop is kept because at the sizes described here the difference is not
observable, and the loop keeps the longest-match rule visible in the code.
Median of repeated runs, relative to each column's own three-pair baseline, so
the figures do not depend on the machine. The three-pair run takes about a
quarter of a second. Neither alternative is quite flat, because the larger
configurations in this benchmark also use longer keys, which cost slightly more
to match.
The regex is only equivalent because `prepare_pairs` already orders patterns
longest first: Python's `re` is leftmost-first, not leftmost-longest.
The regex alternation joins every `value1` into one pattern (`a|bb|c`) that
Python's `re` module scans in C: a constant-factor win, the same work done
faster. The prefix tree is the algorithmic one, taking the pair count out of the
complexity rather than making each step cheaper. The explicit loop is kept
because at the sizes described here the difference is not observable, and it
keeps the longest-match rule visible in the code.
The alternation is only equivalent because `prepare_pairs` orders the patterns
longest first: `re` is leftmost-first rather than leftmost-longest, so `b|bb`
would otherwise match `b` against the input `bb`.
Memory is proportional to the input size, since every line is held before
sorting.
@@ -84,12 +94,13 @@ sorting.
## Error handling
The configuration file is validated while loading: a line without `=`, and a
line whose value1 is empty, are both rejected reporting the offending line
number. An empty value1 would match at every position without consuming
anything, which is why it is refused rather than ignored. Missing files,
unreadable files, and input that is not valid UTF-8 are reported as messages on
stderr rather than tracebacks, and exit with status 1. Status 2 is reserved by
`argparse` for command line misuse.
line whose `value1` is empty, are both rejected, reporting the offending line
number. An empty `value1` would match at every position without consuming
anything, which is why it is refused rather than ignored.
Missing files, unreadable files, and input that is not valid UTF-8 are reported
as messages on stderr rather than tracebacks, and exit with status 1. Status 2
is reserved by `argparse` for command line misuse.
## Tests
@@ -99,8 +110,9 @@ pytest
```
- `tests/unit/` covers configuration parsing and the replacement algorithm,
including the overlapping-pattern and empty-input edge cases, plus a line by
line check of the worked example in `test_sample_data.py`.
including the overlapping-pattern case, empty input, and the rejection of an
empty `value1`, plus a line by line check of the worked example in
`test_sample_data.py`.
- `tests/integration/test_cli.py` invokes `main.py` as a subprocess to cover
what unit tests cannot: argument handling, exit status, stderr messages, and
the sort order as actually printed.