% comby -match-only 'func (:[name] :[type]) :[[method]](:[?args]) :[ret~[\w\s,()]*] {:[before]}' '' .go
% comby -match-only 'func (:[name] :[type]) :[[method]](:[?args]) :[ret] {:[before]}' '' .go
/home/feofan/go/src/playground/at-2020-06-22-124608/snippet.go:25:func (v Vertex) Print() {\n fmt.Printf("%+v\n", v)\n}\n\nfunc (v *Vertex) Scale(f float64) {\n v.X = v.X * f\n v.Y = v.Y * f\n v.Z = v.Z * f\n}==> Starting build()...
Nothing to do.
==> Starting check()...
patdiff (internal) (exit 1)
(cd _build/default && /home/feofan/.opam/4.09.1/bin/patdiff -keep-whitespace -location-style omake -unrefined test/common/test_pipeline.ml test/common/test_pipeline.ml.corrected)
------ test/common/test_pipeline.ml
++++++ test/common/test_pipeline.ml.corrected
File "test/common/test_pipeline.ml", line 37, characters 0-1:
| ; substitute_in_place = true
| ; disable_substring_matching = false
| }
| ; output_printer = (fun _ -> ())
| ; interactive_review = None
| }
|
|let%expect_test "interactive_paths" =
| let _, count =
| let scheduler = Scheduler.create ~number_of_workers:0 () in
| Pipeline.with_scheduler scheduler ~f:(
| Pipeline.process_paths_for_interactive
| ~sequential:false
| ~f:(fun ~input: _ ~path:_ -> (None, 0)) [])
| in
| print_string (Format.sprintf "%d" count);
-| [%expect_exact {|0|}]
+| [%expect.unreachable]
+|[@@expect.uncaught_exn {|
+| (* CR expect_test_collector: This test expectation appears to contain a backtrace.
+| This is strongly discouraged as backtraces are fragile.
+| Please change this test to not include a backtrace. *)
+|
+| (Division_by_zero)
+| Raised at file "src/procs/hack_bucket.ml", line 40, characters 16-51
+| Called from file "src/procs/hack_bucket.ml", line 49, characters 14-41
+| Called from file "src/interface/scheduler.ml", line 59, characters 10-59
+| Called from file "lib/pipeline/pipeline.ml", line 268, characters 6-17
+| Called from file "lib/pipeline/pipeline.ml", line 260, characters 15-26
+| Called from file "lib/pipeline/pipeline.ml", line 260, characters 15-26
+| Called from file "test/common/test_pipeline.ml", line 31, characters 4-165
+| Called from file "collector/expect_test_collector.ml", line 225, characters 12-19 |}]
|
|let%expect_test "launch_editor" =
| let configuration =
| { configuration
| with interactive_review =
| Some
| { editor = "vim"
| ; default_is_accept = true
| }
| }
| in
| let result =
| try Pipeline.run matcher configuration; "passed"
| with _exc -> "Not a tty"
| in
| print_string result;
make: *** [Makefile:28: test] Error 1
==> ERROR: A failure occurred in check().
Aborting...
@s-kostyaev you'll have to remove the space space between :[ret...] and {:[before]}, so that it is just :[ret...]{:[before]}. This is because otherwise comby will try match "space" then then the regex and then space again, which doesn't match the case where there is only one space. Sorry that's not intuitive, that's part of what I'm trying to make easier.
For the crash: ah, I have to rebase the branch to make those tests pass. It should run OK without the passing checks though.
∗ installed shell.v0.12.0
∗ installed patdiff.git
Done.
<><> lwt.4.5.0 installed successfully <><><><><><><><><><><><><><><><><><><><><>
=> Lwt 5.0.0 will make some breaking changes in December 2019. See
https://github.com/ocsigen/lwt/issues/584
<><> bisect_ppx.git installed successfully ><><><><><><><><><><><><><><><><><><>
=> Bisect_ppx 2.0.0 has deprecated some command-line options. See
https://github.com/aantron/bisect_ppx/releases/tag/2.0.0
File "lib/rewriter/rewrite_template.ml", line 22, characters 4-12:
22 | | Ok label -> List.find_map label ~f:ident
^^^^^^^^
Error: This pattern matches values of type
('a, 'b) Core._result = ('a, 'b) Stdlib.result
but a pattern was expected which matches values of type
consume:Angstrom.Consume.t ->
(string option list, string) Stdlib.result
make: *** [Makefile:10: release] Error 1
==> ERROR: A failure occurred in build().
Aborting...comby function '' test.js -o (to match the function keyword for test.js), it instead searches the current directory and fails on an unreadable subdirectory
-stdin and -stdout options. For multiple files combine that with xargs. It should be the case that an absolute or complete relative path like ./path/to/test.js doesn't cause comby to traverse the directory, but in the case that's not working, the piping will definitely work.
I'm trying to use comby for a replacement like this: GetCrew(plr, idx) to GetPlayer(plr)->GetCrew(idx). I came up with comby 'GetCrew(:[plr],:[? ]:[idx])' 'GetPlayer(:[plr])->GetCrew(:[idx])' and have two questions:
1 - is :[? ] the best way to handle optional whitespace between arguments? The examples (e.g., the swap one on the installation page) just seem to assume that there is always a single space. Have you considered a way to strip whitespace in the replacement pattern?
2 - The idx parameter is actually optional, and I'm not sure how to handle that with comby. So I also need to replace GetCrew(plr) with GetPlayer(plr)->GetCrew()
Hey @lluchs :-) Good questions. For the first: since comby runs on the concrete syntax, the only way to specify an optional match in a single template is with the option you used :[? ]. Very recently, I've added regex support to comby, that you can use instead, which is perhaps more familiar and flexible for scanner/lexer kinds of matches. Unfortunately I haven't documented it on the site yet (soon!), but the basic syntax is :[thing~regex], where thing is an optional identifier name. Example on your code: https://bit.ly/315pe70. You'll need v0.18.1, I'm not sure that it is in brew yet, it may be. So, this is the best way I can see to do this, if you want to just use a single template. But:
Have you considered a way to strip whitespace in the replacement pattern?
If you can remove the constraint of specifying just a single template, you can strip whitespace using a rewrite rule. This example will strip the leading whitespace off of everything that matches <spaces>:[idx]: https://bit.ly/2Ycbdm2
For question 2: There's always the possibility of doing this with two invocations. The single argument case would match something like GetCrew(:[[plr]]). There is a way to do everything in one invocation, but it's more complicated. So that said, it turned out to be a really interesting thing to try do with comby. Normally, rewrite rules can get you pretty far on their own, and that's what I tried at first: https://bit.ly/3aBSepX. The problem there is that it's not possible to assign a value to idx in a rule (and that's a need that came up in the past, but not often). It is something I may add, but it lead me to try solve this a different way.
I came up with a way to use regex to match the first argument (assuming it is alphanumeric) , and then a rewrite rule to strip the leading commas/whitespace if it matches the idx part. It looks like this: https://bit.ly/3aDzKoU. Just expand the box to see the rules but here's a screenshot:
:[~[^,]+] to match everything up to the first comma, but of course that then also matches the closing ) if there is no comma, and the whole match fails.
but of course that then also matches the closing ) if there is no comma
Yeah, that's unfortunately the main pitfall of introducing regex. And yep, I assume that if you exclude ) in the regex, like ~[^,)]+ it doesn't work for your use case because the first argument might be an expression like foo->bar() which will then also not (in general) match what you want.
There is a way to implement the notion of "match a balanced expression but exclude <character set>, like comma", instead of "match a regex but exclude <character set>" which would address this. I'd need to introduce a syntax for that, and haven't thought about it yet :-)
Array.from(x).map(z) to Array.from(x, z) but not when Array.from(x, y).map(z) as that would result in Array.from(x, y, z) which would be incorrect (y and z would have to be combined through function composition). :slight_smile:
Yeah, that makes a lot of sense. I've played around with something that'll match syntax for expressions like foo(x, y), so that Array.from(foo(x, y)) still matches despite the whitespace. The tricky part is also supporting something like:
Array.from(function (...) {...})
The function is one expression, but comby doesn't know about function being meaningful/delineating a keyword for the start of a function expression, so will just treat function as token, and the rest as structured blocks. You could use multiple templates here but obvi it's nice to have just hole syntax to match a JS expression (otherwise we wouldn't be talking about this :P). My thinking at this stage is to encode expression holes on a per-language basis so that this just works out of box.
function ... if it spans multiple lines. In which case it starts to make sense to define the syntax (subset) of expressions to match, possibly in a global form for all the patterns.
It involves matching a sub expression that may or may not be nested at different levels of
{} delimiters. Is there any way to match no matter how many levels the sub expression might be?
Hey @jasonkeene glad it's working for you. Right, so by default when the template contains {...} and then a pattern inside, it will only match at that top level of {...}. The only way right now to match a pattern inside {...} in a way that doesn't care about the level of matching is to use a match or rewrite rule. Here's what I think you're aiming for:
func Test:[a](:[b] *testing.T):[c] {:[body]}with a rewrite rule:
where rewrite :[body] {
"if x == y { return:[e] }" -> "skipCond(:[b])"
}Link: https://bit.ly/35jIrUL
When a rule fires on some matched content like :[body], it doesn't care about the level of nesting (it's as if that content is extracted and then running comby only on that part). A side effect here is that you don't need to :[d] or :[f] parts which were basically 'anything' in the previous template--and also, the previous template would stop matching at the first if x == y as well, while the rule will continue and match all instances.
This use case comes up often enough that it makes sense for me to introduce an inline notation to express 'match this expression inside these {...}at any level' so you don't need to write out a separate rule. Related tools like Coccinelle use a kind of notation like <... stuff ...> to express this, and I'm thinking about something similar in comby, something like, say,
func func Test:[a](:[b] *testing.T):[c] {
<... if x == y { return:[e] } ...>
}where this syntax just means that the rule above is created for you :-) Still TBD
Hello, looks like comby 0.18 introduced a breaking change, but I guess this was intended . The simplify of matching for :[hole]: is causing different outputs when using comby 0.15.0 vs. 0.18.3.
Given the follwoing input file
<button>Single line</button>
<button>
Multi-line
</button>You can use this bash script to compare output of comby 0.15.0 and 0.18.3.
COMBY_M="<button>:[hole]</button>"
COMBY_R="<button>###:[hole]###</button>"
echo 'Output comby-0.15.0\n'
comby-0.15.0 "$COMBY_M" "$COMBY_R" ./test.html -stdout
echo '\n---------------------\nOutput comby-0.18.3\n'
comby-0.18.3 "$COMBY_M" "$COMBY_R" ./test.html -stdoutBoth comby calls should produce the same output, but 0.18.3 behaves different which seems to be related to the change introduced in 0.18.0.
Output comby-0.15.0
<button>###Single line###</button>
<button>###
Multi-line
###</button>
---------------------
Output comby-0.18.3
<button>###Single line###</button>
<button>
Multi-line
</button>I wonder how I can replicate the same output with a single match template.
@stefanbuck Ah, you're right! Shoot, it wasn't intended to be a breaking change, I (wrongly) assumed that the single/multiline change would preserve all previous behaviors. I'm going to revert this change and make the new behavior available under a flag.
Separately, I'll then bump the major version where this flag/behavior is enabled by default, since I do think enabling it is a better default behavior for the vast majority of use cases. The new default behavior won't better for your HTML case yet (in which case, the flag will need to be turned off if you decide to use the new major version), however I will make it work by default for the XML/HTML case at a later point, it just needs some bigger changes.
Thanks for feedback!