top | item 31413533

Comby - Structural code search and replace for every language

196 points| WallyFunk | 3 years ago |comby.dev | reply

24 comments

order
[+] feanaro|3 years ago|reply
A more interesting comparison than with Coccinelle is with semgrep (https://semgrep.dev/) since it also supports semantic/structural pattern matching over a number of languages, but also adds in a bunch of Boolean logic into the mix, as well as some more advanced semantic features like constant propagation.

Curious whether the teams behind these tools are aware of each other? I suppose it's unlikely that they're not.

[+] carlmr|3 years ago|reply
Comby works with C++, which to me is a great advantage.
[+] LesZedCB|3 years ago|reply
yeah i just discovered semgrep a week ago! i'll have to download this and compare.

i found the CLI features lacking in semgrep (vs using a yaml config), so maybe this will have more support there.

[+] kretaceous|3 years ago|reply
Okay wow. I didn't tools like this existed! There's been more than once where I've found myself in a situation like "find all code like this, change if this, delete if this". Regex can only get you so far.

Does someone know if this or something similar is available as a VSCode extension? It'd be fun to make one if not!

[+] vanesa-|3 years ago|reply
Sourcegraph Community Manager here. You can use the Sourcegraph extension for VSCode [1] that would allow you to have the structural search powered by Comby (or literal or regular expression search) inside of VSCode.

You can't do batch changes[2] inside of VSCode yet, only via the Sourcegraph CLI[3]. We're working on expanding this.

[1] https://marketplace.visualstudio.com/items?itemName=sourcegr... [2] https://docs.sourcegraph.com/batch_changes/explanations/intr... [3] https://docs.sourcegraph.com/cli

[+] ducktective|3 years ago|reply
language-server methods like rename
[+] adamesque|3 years ago|reply
I believe Sourcegraph uses Comby under the hood to an extent for structural matching as well!
[+] vanesa-|3 years ago|reply
Yep! (Sourcegraph Community Manager here) We use Comby syntax for structural search: https://docs.sourcegraph.com/code_search/reference/structura...

We also have a few tutorials to make large-scale code changes across many repositories and code hosts (batch changes) with Comby:

- Refactor Go code using Comby: https://docs.sourcegraph.com/batch_changes/tutorials/refacto...

- Updating Go import statements using Comby: https://docs.sourcegraph.com/batch_changes/tutorials/updatin...

The author of Comby is a researcher at Sourcegraph building our search platform.

[+] picardo|3 years ago|reply
Superb tool. I used Comby to create a massive (3000+ changes) PR in which I ported JS type annotations into TS typings.
[+] slightknack|3 years ago|reply
I mocked up a prototype for something similar to this using TreeSitter and its query language the other day for some heavy refactoring I had to do. This looks like a really cool project, will have to try it out!
[+] CrazyPyroLinux|3 years ago|reply
This looks a lot like https://semgrep.dev/ Interesting that they are both largely OCaml.
[+] jjice|3 years ago|reply
ML style languages handle tree structures and traversal so well, I'm wondering if that's a large reason for using it.
[+] marcodiego|3 years ago|reply
License?

Also, how does it compare to Coccinelle?

[+] smallnix|3 years ago|reply
License seems to be apache V2, see https://github.com/comby-tools/comby/blob/master/LICENSE.

They have a comparison section in their faq (https://comby.dev/docs/faq):

"Coccinelle uses a declarative syntax for matching constructs like function calls, if-statements, expression blocks, and so on for the C language. Some notable differences are that Coccinelle provides a patch-like format to express transformations (handy when you want to inline smaller changes in a broader context) and metavariable declarations (handy for developing and organizing more sophisticated patterns). For an example, see the file null_ref.cocci. Coccinelle works only for the C language, and may struggle to parse files that contain non-standard C constructs, like GCC inline assembly. Recent Coccinelle work has started expanding support for Java. In contrast, Comby’s syntax targets a wide range of languages and is more robust to matching patterns in the presence of unrecognized constructs. Comby is more suited to writing quick find-replace patterns for languages beyond C, and works well on C languages too."

[+] jedisct1|3 years ago|reply
No support for Zig :(
[+] psanford|3 years ago|reply
Odd, since the page linked says:

> Dedicated parsers refine how Comby understands code, but it also has basic out-of-the-box support for data formats like JSON, recent languages like Zig, or your own DSLs!