comby -config . -f .js. And there is a file named index.js in the current directory.
Hey @volkanunsal! Are you using a configuration file like this: https://comby.dev/docs/configuration
or a directory layout like this: https://comby.dev/docs/cheat-sheet#run-multiple-search-and-replace-templates
For configuration files, you'll have to point to -config ./your-file.toml. For directories, the directory (in this case, the current one) would need to contain a file called match with the pattern.
comby -config comby.toml -f .jsx it seems to be working. :)
setScore(:[1]) and then take different actions based on whether [:1 is a string literal, an integer, a decimal, or an expression. I see that pattern matching can use text or regex. Is there a way to dig a little deeper into comby's type of the match?
setScore(4/3) to evaluate to true, but it matches against | ":[_~\\d+]" -> false because I guess it's matching against the prefix. Not sure how to get around that.
Hey @montyz . Casing out on syntax, like you're doing, is the only way to achieve this right now. Using a program's actual type information is something I've wanted to incorporate for a while, but depends a lot on the language. I'm curious what language you are working in, it might be a good fit for combining type info that way.
For the rules, I think what you came up with is likely the closest you can get right now. The trouble you're having is that match will try find that pattern anywhere in the string (and like you say, matches the prefix). To control that, a helpful answer I want to give you is that you can use regex anchors ^...$ to match whether something is only digits (i.e., integer) and everything else should be checked (which seems to be what you're trying to do?). Unfortunately, that's missing from comby, but I coded up a fix here, and the rule becomes a lot simpler: https://github.com/comby-tools/comby/commit/cbdeaf8e1b4bb86557d6d60418568c0a3f1aca77#diff-f7592766e6cb7ff52ef4395d4e49389aR276-R279
-exclude flag. e.g.comby -config comby.toml -f .jsx -i -exclude '!*.comby.*'
comby in the file name.
That's probably best for now. -exclude '!...' sounds like a double negation so what you're probably really after is a way to include file paths like .*comby.*\.jsx? Piping results of a command like :
find . -name "*comby*.jsx"is maybe another way. But yeah, there's no flag to do this only with comby right now
a.pipe(b, c) to pipe(a, b, c). This should include versions like this.a.pipe(b, c) and this.a().pipe(b, c). It should not include versions like const object = { ...pipe(a, b, c) }.Can anyone recommend a way to match this precisely? I thought
:[a.].pipe(:[b]) would do it, but it doesn't match new lines as I would want it to, and it incorrectly matches the object spread example I gave.
pipe
Hey @OliverJAsh. Good question. Your intuition about :[a.] is good. The reason that doesn't work is because it doesn't match special syntax like (). There is an undocumented way to get what you want. It's undocumented because I was experimenting with more ways to match constructs like that one, but haven't settled on a suitable syntax yet :-).
You can do so by changing :[a.] to :[a:e]: https://bit.ly/36JpYC6. You can think of the :e syntax meaning ' match an expression-like element'. It matches contiguous sequences of (...), {...}, [...] as well as characters matched by :[a.].
Awesome thank you @rvantonder. Do you have any ideas how I can get it to match new lines as well? E.g.
this
.a()
.pipe(b, c)@OliverJAsh there isn't a very elegant way to deal with negating { matching-pattern } but in general you may be able to get away with using regex. Here's an example where you optionally match on leading or trailing { }s around your expression, and can use a rule to filter the match based on whether, say, start matched {, or the empty string: https://bit.ly/36OCuQD. Whether this works for you depends on whether that assumption is strong enough to filter out unwanted matches.
I think it's real tricky to deal with matching that pattern over newlines. You can certainly tweak the matching to include newlines, but the problem is that it's difficult to distinguish statements like these in Comby:
first.statement()
second.statement()
.across()
.newline(), because it doesn't know enough about the JS grammar to understand the rules for sequencing with . and how it can spread over newlines/whitespace. The way I came up with would be to deal with this syntax in its own way, but I'm not sure that it'll work for you and it's crude. The idea is basically to say "OK, a chain can be something like an identifier followed by one or more lines that perform an access like .foo that must be prefixed by <whitespace><dot>". Once all that syntax is matched, I use a rule to rewrite the ones that call pipe https://bit.ly/3dg2mGs. I don't know if this works in general, it's just my attempt to deal with multi-line chains like this, and Comby might just not be a good fit for this problem until there's a way to teach it more about JS :-)
Is there a way to do text transformation?
I have an application with hundreds of routes that look like this:
app.get("/foo/bar", auth, (req, res) => {
// ... do stuff
})I want to convert them to something like this:
app.get("/foo/bar", auth, getFooBar);
function getFooBar (req, res) {
// ... do stuff
}@coolaj86 nice idea. Unfortunately there's no builtin for capitalizing strings/letters. That's something worth mentioning in the docs, so I'll update it. Capitalizing strings came up before a handful of times, so it might be something worth adding.
Anyway, right now there is only a workaround using a rewrite rule--it works, but it's a bit silly :-) Assuming your routes are alphanumeric, you can rewrite parts of a string like /f to F for each letter. Here's a short version for getFooBar so you get the idea:
where
match :[path] {| ":[fn]" -> true },
rewrite :[fn] { "/f:[[w]]" -> "F:[w]" },
rewrite :[fn] { "/b:[[w]]" -> "B:[w]" }https://bit.ly/362jA6I (expand the where... box.)
I went ahead and made you a config file for your refactor with a rule that translates all the letters--I might just make a rule like this a builtin somewhere :-) See https://gist.github.com/rvantonder/01b930f7e5d081a5c84f6074fbf3bd13
(FYI I generated the tedious part of the rule using comby https://bit.ly/323bNo4)
@rvantonder I found some bugs (Ithink), but I don't know if they're worth fixing:
There was some code like this
app.get('/*', foo)and it seemed to treat the string '/*' as a comment block.
There was also a regex that it failed to ignore, it was something like this:
app.get(/\(.*\)/, foo)and it seems to have interpreted the stuff in the regex as a closing ).
I'll find out what it actually is and post the example.
I've been using XmlStarlet to swap dependencies in maven pom.xml-files for things like javax->jakarta migrations, but xmlStarlet is limited to swapping one value at a time. In comes comby. I've been trying (and failing to write a template files that will match the xml for my dependencies and only swap the values for matched tags, leaving whitespace intact, not to confuse (or upset) the teams that will be receiving my mountain of PRs in the end. Here bit.ly/2UtSro6 is one of my attempts, that might shed some light on what im trying to do. And here bit.ly/3eX3mQB is a naive attempt, that does not respect original indentation style.
Hey @jensim, I can think of two ways to go about this. I think the easiest might be to do your changes and re-format the files for the appropriate indentation (I think you have lots of options here with existing tools).
There's probably a way to get comby to rewrite it while preserving whitespace, but it means you have to match carefully on tags or blocks to dance around rewriting things that could erase whitespace, here's some examples that might help your case: https://gist.github.com/rvantonder/324a4f1c3a46b15628b72745679bf821
5 replies
I'm not sure how far you can go with it
Hey @bentxt. I'm not really sure either :-) I think generating bindings could work, it all depends on how much you need to take into account for the use case (e.g., do the bindings need type info? something else?). You could use it for macros--I'm not sure that I would embed it in a language, but would work well for general search-replace things.
To execute a sequence of match/rewrites you could use a config file: https://comby.dev/docs/configuration#toml-format
Why do you have doubts about using as a macro (code transformation) tool?
I guess it depends where it fits into a toolchain :-) I think it fits well into a preprocessing script, standalone find-replace, or a custom template generation tool. If I were designing a language from scratch with a macro system though, I don't think it's best to, for example, use comby as an external dependency for implementing that macro system--in this case I would rather (eventually) write the macro system for such a language because it would probably be more powerful with language-specific taken into account. But comby could probably work well to prototype things at a very early stage.
Is it possible to recursively transform captured sections ('body' in this case)?
If I follow you correctly, sure, have a look at https://comby.dev/docs/advanced-usage#rewrite-expressions. rewrite expressions cannot nest themselves, but usually it's possible to do multiple rewrites to match on a subpart to change. Let me know if you have a specific example.
-match-newline-at-toplevelto your CLI command, you should get the same matching behavior as comby.live.