001 - ShellCheck Integration
| Status | accepted |
| Date | 2024-08-29 |
| Decider | timo-reymann |
ShellCheck Integration
Context and Problem Statement
To make the CLI tool useful and support shell script linting it is necessary to integrate shellcheck which is quite mature and covers a lot of ground.
Decision Drivers
- integration is straightforward
- embeddable into the application (no dependencies to install for the user)
Considered Options
- Link the statically compiled library from shellcheck
- Use memexec and embed the binary directly
Decision Outcome
Chosen option: "Use memexec and embed the binary directly," because it does not require to write any haskell or c code.
Pros and Cons of the Options
Link the statically compiled library from shellcheck
Shellcheck is written in Haskell and built via cabal-install.
Fortunately the build can produce a static library libshellcheck.a for all supported platforms.
- Good, because there is no hacky memexec workaround and the memory is shared with the application
- Good, because no need to parse the output as the data structures can be directly used
- Bad, because the libraries are not part of an release so they have to be built manually which takes some time and resources
- Bad, because that makes it harder to embed when done dynamically and when done statically requires a dedicated project to provide them
Use memexec and embed the binary directly
On Linux & macOS it is pretty convenient to make a binary from memory executable and treat it like a regular file. For Windows the workaround is to copy it to a temporary folder on usage and deleting it afterward.
Fortunately for Go, there exists a library that does exactly this.
- Good, because it is super easy to integrate as the binary can be included as is
- Good, because binaries can be directly copied to the repository, making it easy to build for everyone
- Good, because there is no need to use the library and interact with Haskell over CGO which allows static compilation
- Good, because the CLI interface is considerable more stable than the internal library
- Bad, because version updates need to be done manually by copying files
- Bad, because the output needs to be parsed again to be usable
- Bad, because it increases memory usage