follow #15357 and move decodeQuery

[ringabout]

decodeData is very common to use and should work in JS backend(cgi doesn't work in JS backend).

So I move decodeData to std/uri and rename it to decodeQuery. Move CgiError etc to std/uri to keep backwards compatibility.

Or we should continue #15357 and change proc to iterator.

[timotheecour]

Change it to iterator and make cgi call this

[Araq]

Sorry but this seems to be a bad idea.

"Added decodeQuery to std/uri and deprecated decodeData in std/cgi."

Omg why? It's std/cgi, don't deprecate things in there. Stop messing around with the details -- some people here try to write code against the std libraries...

[ringabout]

So we tend not to deprecate duplicated things in stdlib and only deprecate wrong/outdated things in stdlib?

[Araq]

Your code introduced the duplication as far as I can tell. We should avoid both the duplication and the deprecation process.

[timotheecour]

Your code introduced the duplication as far as I can tell. We should avoid both the duplication and the deprecation process.

as of right now, only the API declaration is duplicated between cgi and uri, not the implementation, that's an acceptable tradeoff vs the benefits it brings:

decodeData is very common to use and should work in JS backend(cgi doesn't work in JS backend).

plus discoverability, decodeQuery really does belong in std/uri. And this isn't the 1st PR attempting to address this.

So the current version of this PR is fine IMO, not deprecating cgi.decodeData (but uri.decodeQuery could be mentioned in docs for cgi.decodeData) and introducing uri.decodeQuery

[ringabout]

ECgi* = object of EIO  ## the exception that is raised, if a CGI error occurs

ECgi has changed to CgiError, but some docs didn't change. I will clean them after this PR.

Does decodeQuery need since? I try to add since(1, 5, 1) but failed in CI. It told me that

/home/runner/work/Nim/Nim/lib/pure/cgi.nim(88, 29) Error: undeclared identifier: 'decodeQuery'
FAILURE

[timotheecour]

Does decodeQuery need since? I try to add since(1, 5, 1) but failed in CI. It told me that

probably bootstrap but can't reproduce locally (bin/nim_csources c koch works for me); can you provide a full link to CI logs? I can see several ways to handle this, including when Nim >= (1,5,1) or Nim == (0,20,0) (in pseudocode), depending on the CI log error

[ringabout]

This one:
b43dbc9
And:
https://github.com/nim-lang/Nim/runs/1374923254

[timotheecour]

ok, I can reproduce (locally too, via rm bin/nim and booting from csources); this would be easily fixed via {.privateImport.} (see #11865):

# uri.nim:
iterator decodeQuery(data: string): tuple[key, value: TaintedString] = ...
since (1,5,1): export decodeQuery

# cgi.nim
from uri {.privateImport.} import decodeQuery

(or equivalently, via iterator decodeQueryImpl(data: string): tuple[key, value: TaintedString] = ...; template decodeQuery*(data: string) {.since: (1,5,1).}: untyped = decodeQueryImpl(data)
and in cgi.nim: from uri {.privateImport.} import decodeQueryImpl)

but until then the simplest IMO is to just leave out since in this case.
EDIT: I've rebased #11865 so this option can still be on the table if needed (now or later)

[timotheecour]

needs rebase against devel for conflict