use preprocessor to map define'd constants/macros about nimterop HOT 10 CLOSED

I believe this is now implemented with -dD. We need to add tests though.

from nimterop.

This seems as appropriate an issue thread as any to chime in, but I saw that I think you (shashlick?) in IRC a few days ago asked about cligen handling -I multiple times or probably similarly for -D. No one seemed to answer you, and I suppose there's a chance you tried it and saw that it "just worked", but its CL usage is a little subtle.

An explicit = or : or long option param separation causes things to "reset" to the argument. So, --defines=def0,def9 will clobber any existing seq with just ["def0", "def9"]. A short option (if you added short={ "defines": 'D', "includeDirs", 'I' } ) would let the CLI user say -Ddef1 -Ddef2 with the intended accumulating meaning. They can also use an explicit append mode via --defines+=def3 and even a prepend mode -I^=earlierpath. I found it a little tricky balancing the "need to reset" with the principle of least surprise, and this is only really in the documentation for argcvt. I could be convinced that maybe "reset and assign" should be spelled .= or something. There's been talk of the need for a --help-syntax or something more to guide CL users.

If you hate this syntax, you can also override all of this default syntax by just by doing an import cligen, cligen/argcvt and, between that import and the dispatch. defining your own version of proc argParse*[T](dst: var seq[T], ... . It's only a 19 line proc in cligen/argcvt.nim, but you might also need to learn some more internals to do it.

from nimterop.

@c-blake - thanks for chiming in. I wasn't able to since the += and ^= syntax isn't obvious and there weren't any examples either. I also just tried to see the argcvt docs and it doesn't seem to work. Says table not accessible.

That being said, @timotheecour has already migrated nimterop to use cligen. Here's the syntax so far:

Usage:
  main [required&optional-params]
  Options(opt-arg sep :|=|spc):
  -h, --help                                  write this help to stdout
  -m=, --mode=         string       "cpp"     language; see CompileMode
  -p, --past           bool         false     print AST output
  --pnim               bool         false     run preprocessor on header
  --pretty             bool         true      set pretty
  --preprocess         bool         false     print Nim output
  -d=, --defines=      ,SV[string]  EMPTY     definitions to pass to preprocessor
  -i=, --includeDirs=  ,SV[string]  EMPTY     include directory to pass to preprocessor
  -s=, --source=       string       REQUIRED  C/C++ source/header

from nimterop.

Not sure what's wrong with the http://htmlpreview.github.io/?, but I also get that error a day later. That file is just the output of nim doc argcvt.nim. So, you can also just read the doc comments there.

Also, I don't think I was so clear...Had you tried just "-i dir1 -i dir2" (with no "=") then it would have appended rather than doing a clobbering assignment. It only does the clobbering assignment with an explicit '=' (at least right now). So, it seemed possible to me you might have discovered that subtle property by accident. I don't think of += as very likely to be tried. But, yeah, there should be example usage/--help-syntax. parsopt3 is also quite a bit more flexible than the stdlib parseopts.

from nimterop.

Not sure what's wrong with the http://htmlpreview.github.io/?, but I also get that error a day later.

not sure what you meant above but http://htmlpreview.github.io/?https://github.com/c-blake/cligen/blob/master/argcvt.html worked for me (link given at end of README in https://github.com/c-blake/cligen ; admittedly I'd prefer if that link appeared on top of README)

--help-syntax

=> moved that discussion to c-blake/cligen#64

from nimterop.

@genotrance

I believe this is now implemented with -dD. We need to add tests though.

is it, though? that's why I was trying to test with #17 but it fails

from nimterop.

#17 was last tested 5 days ago but preprocessor was only enabled 4 days ago. Maybe if we close and reopen it, it might pass.

7344042

from nimterop.

indeed, travis now passes! appveyor still fails though

EDIT: hmm, appveyor fails on ubuntu...

/home/appveyor/projects/nimterop/tests/tnimterop_c.nim(21, 15): Check failed: myvar1 == 20
myvar1 was 10
Error: execution of an external program failed: '/home/appveyor/projects/nimterop/tests/tnimterop_c '
stack trace: (most recent call last)
/home/appveyor/projects/nimterop/nimterop.nimble(20) testTask
/home/appveyor/projects/nimterop/nimterop.nimble(17) execCmd
/home/appveyor/binaries/nim-0.19.0/lib/system/nimscript.nim(237) exec
/home/appveyor/binaries/nim-0.19.0/lib/system/nimscript.nim(237, 7) Error: unhandled exception: FAILED: nim c -r tests/tnimterop_c.nim

from nimterop.

This has been fixed.

from nimterop.

@genotrance could we reopen?
constants works but not C macros, see #30

EDIT: moved to #31

from nimterop.