paths-le 0.2.2

Extract every file path in a codebase, and say whether it still points at anything
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108

[package]
name = "paths-le"
version = "0.2.2"
edition = "2024"
rust-version = "1.88"
license = "MIT"
readme = "README.md"
description = "Extract every file path in a codebase, and say whether it still points at anything"
authors = ["Nolin D Naidoo"]
homepage = "https://letools.dev/tools/paths-le"
repository = "https://github.com/nolindnaidoo/paths-le"
keywords = ["paths", "broken-links", "symlink", "filesystem", "link-checker"]
categories = ["command-line-utilities", "filesystem", "development-tools", "text-processing", "parsing"]
# What a consumer gets, and why.
#
# Ships: src/, Cargo.toml, Cargo.lock, LICENSE, README.md — needed to
# build the binary. SPEC.md and CHANGELOG.md — the behavioural contract
# someone scripting against the exit codes needs. tests/ and fixtures/ —
# not needed to *build* (verified: the crate compiles without them), but
# needed to *verify*: `cargo test` on the unpacked tarball runs the full
# corpus, so the parity claims in the README are checkable by whoever
# installed it rather than taken on trust.
#
# Excluded: AGENTS.md and CLAUDE.md are how this repo is worked on, not
# how the tool is used. rust-toolchain.toml is a convenience for
# contributors to *this* checkout; shipped, it would pin the toolchain
# of anyone who unpacks or vendors the crate, which is not this crate's
# business.
exclude = ["AGENTS.md", "CLAUDE.md", "rust-toolchain.toml"]

[[bin]]
name = "paths-le"
path = "src/main.rs"

[dependencies]
serde = { version = "1", features = ["derive"] }
serde_json = "1"
# Two regex engines, on purpose, and the split is the point.
#
# `regex` is linear-time and its matching cannot fail, so every pattern
# that does not need backtracking uses it and needs no error path.
#
# `fancy-regex` exists for one reason: the extension's quote-matching
# patterns use backreferences — `(['"])([^'"\n]+)\1` — which `regex` does
# not support. Rewriting them as alternations would change the patterns,
# their group numbering, and their behaviour on strings containing the
# other quote character. Porting them verbatim is parity; rewriting them
# is parity drift waiting to happen, so the dependency is cheaper than
# the rewrite. It is pure Rust and already carries `regex` beneath it, so
# naming `regex` directly costs nothing further.
regex = "1"
fancy-regex = "0.19"
# The format parsers, each chosen to match what the extension uses:
# jsonc-parser for jsonc-parser (comments, trailing commas, real
# offsets), toml for @iarna/toml, csv for csv-parse. dotenv, HTML, CSS
# and JavaScript are regex in the extension and regex here.
jsonc-parser = "0.33"
# `preserve_order` is not optional. Without it a TOML table is a
# BTreeMap and iterates alphabetically, so extraction order — and with
# it the forward-moving position locator the extension uses — stops
# following the document. The corpus catches this immediately, which is
# the only reason it is written down here rather than discovered twice.
toml = { version = "1", features = ["preserve_order"] }
csv = "1"
# saphyr for js-yaml, matching what numbers-le pairs them as. A YAML
# document is where the paths in a repository's CI configs, Kubernetes
# manifests and compose files live, and there is no reading one without
# a parser: indentation, anchors, block scalars and flow collections are
# not a regex. Positions still come from a forward-moving text search,
# the way TOML's do, so the two parsers only have to agree on values.
saphyr = "0.0.11"
# The walker. `ignore` is ripgrep's, so "what this tool walks" and "what
# ripgrep walks" are the same answer — which is the answer a person
# auditing a repository already has in their head.
ignore = "0.4"

[profile.release]
strip = true
lto = "thin"
codegen-units = 1
# Every output of this tool is a claim someone scripts against — a line
# number, a column, a count. A wrapped number is silently wrong data in a
# report whose whole value is honesty, which is worse than a crash — so
# overflow keeps panicking in release, as it does in debug. (Same rule as
# pixelcoords, pixelactions and scrape-le.)
overflow-checks = true

[lints.rust]
unsafe_code = "forbid"

[lints.clippy]
all = { level = "warn", priority = -1 }
pedantic = { level = "warn", priority = -1 }
# Deliberate relaxations, same policy as the other crates: no inline
# #[allow] anywhere (CI enforces); anything relaxed must be listed here,
# visibly.
cast_possible_truncation = "allow"
cast_possible_wrap = "allow"
cast_sign_loss = "allow"
cast_precision_loss = "allow"
must_use_candidate = "allow"
missing_errors_doc = "allow"
missing_panics_doc = "allow"
# The parsed command line is a flat record of the flags that were typed,
# and it has more than three of them. Grouping booleans into sub-structs
# to satisfy a lint would put the shape of the type ahead of the shape
# of the interface it mirrors.
struct_excessive_bools = "allow"