Implement a constraint parser for more complex inputs

[tegefaulkes]

Specification

I'm re-opening this. We want to explore using a Constraint parser for parsing the vaults secret path. This should also allow for escaping certain characters we'd otherwise disallow such as =. For example, vault:this/secret\=name\==value should be valid. vault:this/secret\=name==value would be invalid for having two un-escaped = and vault:this/secret\=name\==value= for having an invalid = in the value section.

That said, for the usage of secret env command, an = inside the secret name would be invalid since it would be an invalid environment variable name. Directory paths should be able to include = however.

Right now I'm pegging this expansion to the logic as a low priority. It's a lot of extra work just to allow = in the secret directory names.

On top of this, I think the error for failing to parse the vault secret path could be clearer and conditional. For example, if it failed to parse due to an invalid character, the exact reason needs to be defined.

Additional context

Related #199

Tasks

1. Further spec out a constraint parser for more complex parsing of inputs and options.
2. Apply this constraint parser to parsing secrets paths.

[linear]

ENG-353 Implement a constraint parser for more complex inputs

[CMCDragonkai]

I think your backtick formatting needs a bit of a fix above.

[CMCDragonkai]

This is not that difficult. Use chatgpt and write out a LL1 parser. If we need backtracking, peg parsing works too. This is actually comp sci 101.

[CMCDragonkai]

@aryanjassal

[CMCDragonkai]

Backtracking parsers allow one to write simpler grammars that retain a sort of ambiguity. The resulting grammar is more compact by leveraging ambiguity that's resolved by pattern matches or semantic predicates.

I'm following parsing combinators in my design of the parser utilities.

https://chatgpt.com/share/e/065e9f3d-4bc4-4d5e-9ec4-e95ab864e9a9

They are meant to be simple and easily human debuggable compared to peg or ll(K). If we ever need more sophisticated then peg. However this problem is too simple. It should be solvable with parsing combinators like we are currently doing.

[CMCDragonkai]

If the problem here can be more well specified with clear examples, I can feed it into ChatGPT.