Check regexes at compile

[gabriel-fallen]

Currently regular expressions (TRegex for SchemaText) get compiled only when (and if) relevant data being validated at run-time. Thus even malformed regexes pass compilation and might sleep past testing. It would be nice if they get compiled or syntactically verified while compiling Haskell.

[dredozubov]

It would. But do you have a way of doing that in mind?

[sboosali]

There's not much you can do with type-level strings even in GHC 8.2. There's probably a type checker plug in somewhere, or a macro would work, but those have the obvious drawbacks.

…

On Feb 14, 2018 7:14 AM, "Denis Redozubov" ***@***.***> wrote: It would. But do you have a way of doing that in mind? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACNoMUjD7LdVc2iTI5NEpPv3jwP33UtSks5tUvgsgaJpZM4SFdu4> .

[sboosali]

However, a type-level regex representation, well it's convenient, could work. Like a data kind for each case: data REGEX = LITERAL Symbol | STAR REGEX | ...

…

On Feb 14, 2018 2:11 PM, "Spiros Boosalis" ***@***.***> wrote: There's not much you can do with type-level strings even in GHC 8.2. There's probably a type checker plug in somewhere, or a macro would work, but those have the obvious drawbacks. On Feb 14, 2018 7:14 AM, "Denis Redozubov" ***@***.***> wrote: > It would. But do you have a way of doing that in mind? > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#8 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/ACNoMUjD7LdVc2iTI5NEpPv3jwP33UtSks5tUvgsgaJpZM4SFdu4> > . >

[dredozubov]

I've thought about the type level representation, but I would rather separate something like this to a different project before trying to incorporate it into schematic. Schematic already incurs heavy compilation burden on ghc and I think it can get worse with type level regexes. Do you think some helpers to test out the type level regexes would come in handy? It may be an alternative.

[sboosali]

Yeah, wrapping whatever regular expression parser is eventually invoked makes sense.