Opinionated OCaml linter with Merlin integration for code quality, naming conventions, and style checks
merlint: add E830 inlined-algorithm detection for interop generators
Completes the E8xx interop testing rule series. E830 flags generator
scripts that define encode/decode/compute functions locally instead of
calling the upstream oracle's public API — the core principle of interop
testing.
+180
+49
docs/index.html
+49
docs/index.html
···
1588
1588
</div>
1589
1589
</div></div>
1590
1590
</div>
1591
+
<div class="error-card" id="E800">
1592
+
<div>
1593
+
<span class="error-code">E800</span>
1594
+
<span class="error-title">Missing generate.sh</span>
1595
+
</div>
1596
+
<div class="error-hint"><p>Every interop test must have scripts/generate.sh as the single entry point for trace regeneration via `dune build @regen-traces`.</p></div>
1597
+
</div>
1598
+
<div class="error-card" id="E801">
1599
+
<div>
1600
+
<span class="error-code">E801</span>
1601
+
<span class="error-title">Interop dir named after language</span>
1602
+
</div>
1603
+
<div class="error-hint"><p>Interop test directories should be named after the oracle tool (e.g. spacepackets, dariol83, crcmod), not the language (e.g. python, go). This makes it clear which external implementation is the reference.</p></div>
1604
+
</div>
1605
+
<div class="error-card" id="E805">
1606
+
<div>
1607
+
<span class="error-code">E805</span>
1608
+
<span class="error-title">Missing requirements.txt</span>
1609
+
</div>
1610
+
<div class="error-hint"><p>Python oracles must pin dependencies in requirements.txt with exact versions (e.g. crcmod==1.7). This ensures reproducible trace generation without depending on local installs.</p></div>
1611
+
</div>
1612
+
<div class="error-card" id="E810">
1613
+
<div>
1614
+
<span class="error-code">E810</span>
1615
+
<span class="error-title">Missing regen-traces alias</span>
1616
+
</div>
1617
+
<div class="error-hint"><p>Every interop test dune file must define a regen-traces alias as the single trigger for refreshing traces: `(rule (alias regen-traces) ...)`.</p></div>
1618
+
</div>
1619
+
<div class="error-card" id="E815">
1620
+
<div>
1621
+
<span class="error-code">E815</span>
1622
+
<span class="error-title">REGEN_TRACES sentinel in dune</span>
1623
+
</div>
1624
+
<div class="error-hint"><p>The regen-traces alias should be the single entry point — no REGEN_TRACES=1 env var sentinel. Remove the (enabled_if ...) guard so `dune build @regen-traces` works directly.</p></div>
1625
+
</div>
1626
+
<div class="error-card" id="E820">
1627
+
<div>
1628
+
<span class="error-code">E820</span>
1629
+
<span class="error-title">Hand-rolled CSV parsing</span>
1630
+
</div>
1631
+
<div class="error-hint"><p>Use csvt (Csvt.decode_file with a Csvt.Row codec) for CSV trace parsing. Never hand-roll CSV readers with open_in/input_line/split_on_char.</p></div>
1632
+
</div>
1633
+
<div class="error-card" id="E830">
1634
+
<div>
1635
+
<span class="error-code">E830</span>
1636
+
<span class="error-title">Inlined algorithm in generator</span>
1637
+
</div>
1638
+
<div class="error-hint"><p>The generator MUST call the upstream tool's public API. Never reimplement the algorithm being verified — this defeats the purpose of interop testing. If the public API doesn't expose what you need, drop the test rather than inlining.</p></div>
1639
+
</div>
1591
1640
1592
1641
<a href="#top" class="back-to-top">↑ Top</a>
1593
1642
</body>
+7
lib/data.ml
+7
lib/data.ml
+124
lib/rules/e830.ml
+124
lib/rules/e830.ml
···
1
+
(** E830: Inlined algorithm in interop generator *)
2
+
3
+
type payload = { dir : string; file : string; reason : string }
4
+
5
+
(* Heuristic: a generator that defines functions with algorithmic names
6
+
(encode, decode, compute, calculate, process, transform, convert)
7
+
is likely reimplementing the algorithm instead of calling the oracle. *)
8
+
let suspicious_defs =
9
+
[
10
+
(* Python: def <name>( *)
11
+
(fun line ->
12
+
let l = String.trim line in
13
+
String.length l > 4
14
+
&& String.sub l 0 4 = "def "
15
+
&&
16
+
let name =
17
+
try
18
+
let i = String.index l '(' in
19
+
String.sub l 4 (i - 4) |> String.trim
20
+
with Not_found -> ""
21
+
in
22
+
List.exists
23
+
(fun prefix ->
24
+
Astring.String.is_prefix ~affix:prefix (String.lowercase_ascii name))
25
+
[ "encode"; "decode"; "compute"; "calculate"; "process"; "transform" ]
26
+
&&
27
+
(* Exclude test/verification helpers *)
28
+
not
29
+
(Astring.String.is_prefix ~affix:"test" (String.lowercase_ascii name)
30
+
|| Astring.String.is_prefix ~affix:"verify"
31
+
(String.lowercase_ascii name)));
32
+
(* Rust: fn <name>( or pub fn <name>( *)
33
+
(fun line ->
34
+
let l = String.trim line in
35
+
let after_fn =
36
+
if Astring.String.is_prefix ~affix:"fn " l then
37
+
Some (String.sub l 3 (String.length l - 3))
38
+
else if Astring.String.is_prefix ~affix:"pub fn " l then
39
+
Some (String.sub l 7 (String.length l - 7))
40
+
else None
41
+
in
42
+
match after_fn with
43
+
| None -> false
44
+
| Some rest ->
45
+
let name =
46
+
try
47
+
let i = String.index rest '(' in
48
+
String.sub rest 0 i |> String.trim
49
+
with Not_found -> ""
50
+
in
51
+
List.exists
52
+
(fun prefix -> Astring.String.is_prefix ~affix:prefix name)
53
+
[ "encode"; "decode"; "compute"; "calculate"; "process" ]);
54
+
]
55
+
56
+
let scan_file path =
57
+
try
58
+
let ic = open_in path in
59
+
let found = ref false in
60
+
(try
61
+
while true do
62
+
let line = input_line ic in
63
+
if List.exists (fun check -> check line) suspicious_defs then
64
+
found := true
65
+
done
66
+
with End_of_file -> ());
67
+
close_in ic;
68
+
!found
69
+
with Sys_error _ -> false
70
+
71
+
let check (ctx : Context.project) =
72
+
let dirs = Interop.find_oracle_dirs ctx.project_root in
73
+
List.filter_map
74
+
(fun (d : Interop.oracle_dir) ->
75
+
let scripts = Filename.concat d.path "scripts" in
76
+
if not (Sys.file_exists scripts) then None
77
+
else
78
+
let files =
79
+
try Sys.readdir scripts |> Array.to_list with Sys_error _ -> []
80
+
in
81
+
let generator_files =
82
+
List.filter
83
+
(fun f ->
84
+
let f = String.lowercase_ascii f in
85
+
Astring.String.is_prefix ~affix:"generate" f
86
+
&& (Filename.check_suffix f ".py"
87
+
|| Filename.check_suffix f ".rs"
88
+
|| Filename.check_suffix f ".go"))
89
+
files
90
+
in
91
+
let inlined =
92
+
List.filter_map
93
+
(fun f ->
94
+
let path = Filename.concat scripts f in
95
+
if scan_file path then Some f else None)
96
+
generator_files
97
+
in
98
+
match inlined with
99
+
| [] -> None
100
+
| file :: _ ->
101
+
Some
102
+
(Issue.v
103
+
{
104
+
dir = d.path;
105
+
file;
106
+
reason =
107
+
"defines encode/decode/compute functions — may be \
108
+
reimplementing the algorithm instead of calling the \
109
+
oracle's API";
110
+
}))
111
+
dirs
112
+
113
+
let pp ppf { dir; file; reason } =
114
+
Fmt.pf ppf "Interop generator %s/scripts/%s: %s" dir file reason
115
+
116
+
let rule =
117
+
Rule.v ~code:"E830" ~title:"Inlined algorithm in generator"
118
+
~category:Interop_testing
119
+
~hint:
120
+
"The generator MUST call the upstream tool's public API. Never \
121
+
reimplement the algorithm being verified — this defeats the purpose of \
122
+
interop testing. If the public API doesn't expose what you need, drop \
123
+
the test rather than inlining."
124
+
~examples:[] ~pp (Project check)