add linenoise.readLineStatus to get status (eg: ctrl-D or ctrl-C)

[timotheecour]

• close Linenoise returned nil [IOError] #5795 by providing a suitable workaround; since we're now returning a status indicator, user can forward this to call into exitHandler if status was ctrl-C for example
• allows fixing nim secret being un-quittable, as described in control+D should exit nim secret; quit sometimes doesn't work timotheecour/Nim#533: in a next PR, we can use the new API readLineStatus and exit the nim secret REPL if status was ctrl-D
• allows user to implement a very common shell/REPL feature (found in bash, python, and pretty much all REPL I've used) which is to quit the REPL loop on ctrl-D, but just abort the prompt and go to next prompt on ctrl-C without exiting the loop

note

• another recent linenoise PR was closed see update linenoise #16451 (comment)

Please rewrite all the C code in Nim. We don't need more "static" variables, it's a desaster for multi-threading.

however this PR doesn't add more static variables, insted the state is propagated as a side variable allocated on the stack.

• This is a small diff on the c side and the re-implementing in nim could be handled later; furthermore it's not clear re-implementing in nim is the correct approach, instead we should leverage nimble packages (eg https://github.com/jangko/nim-noise or https://github.com/de-odex/linenoise-nim) as described in update linenoise #16451 (comment)

• I'm using the nimExperimentalFoo pattern described in nimExperimentalFoo for experimental modules/APIs timotheecour/Nim#575

links

• https://fossies.org/linux/txr/linenoise/linenoise.h seems to have a lino_error exposed which has a similar goal, but this is a different port

   40 typedef enum lino_error {
   41     lino_no_error,      /* No error has occurred. */
   42     lino_error,         /* Unspecified error */
   43     lino_eof,           /* Line input terminated by Ctrl-D or EOF. */
   44     lino_ioerr,         /* Line input terminated by I/O error. */
   45     lino_notty,         /* Input is not a terminal. */
   46     lino_intr           /* Line innput terminated by Ctrl-C or interrupt */
   47 } lino_error_t;

• see also distinguish between ctrl C vs ctrl D jangko/nim-noise#18 (similar thing for pkg nim-noise

future work

• make nim secret + friends recognize ctrl-D => ^D now works with readLineFromStdin (in particular in nim secret) #18442

allows fixing nim secret being un-quittable, as described in timotheecour#533: in a next PR, we can use the new API readLineStatus and exit the nim secret REPL if status was ctrl-D

[Araq]

New enum types should be non-pure, non-prefix but Uppercased, but this probably requires an RFC...

[timotheecour]

New enum types should be non-pure, non-prefix but Uppercased, but this probably requires an RFC...

that seems controversial; non-pure and non-prefix, even if Uppercased, increases likelihood of name clashes (because of other enum's or types or const symbols).

New enum types should be non-prefix, {.pure.}, lowercase would make more sense IMO.

EDIT: as done in compilesettings which uses the quite intuitive/readable:

type
  SingleValueSetting* {.pure.} = enum ## \
                      ## settings resulting in a single string value
    arguments,        ## experimental: the arguments passed after '-r'
    outFile,          ## experimental: the output file
    ...

[Araq]

Pure enums have usability problems which is why every language ends up with a mechanism like "let's have a using statement inside switch" (C++). Name clashes should be resolved via EnumType.value if and when they come up, just like every other name clash.