nim jsondoc -o:doc.json filename.nim fails when sequences without a type are used

[faulander]

Running "nim jsondoc -o:doc.json filename.nim" results in:
Error: unhandled exception: index out of bounds, the container is empty [IndexError]

Example

proc hashesToStr(hashes: seq): string = 
  var h:string
  if hashes.len == 0:
      result = ""
  else:
    for hash in hashes:
      h = h & hash & "|"
    h = h[0 .. len(h)-2] # cuts of the last |
    result = h

Additional Information

Already discussed and found a workaround with PMunch on IRC.

$ nim -v
Nim Compiler Version 1.2.0

[metagn]

Unrelated to the issue, and not a workaround, can be marked as offtopic, just an FYI: instead of doing str = str & ... (creates a new string) you should do str &= ... which is actually an alias for str.add(...) (modifies the existing string), and instead of doing str = str[0.. newlen - 1] (creates a copy) you should do str.setLen(newlen) (doesn't create a copy).

[PMunch]

The issue is that it assumes that seq in the signature has a type and fails when it doesn't. Should be a simple fix in compiler/types.nim:598 and it should be fixed similarly for the other types with the same behaviour. I'll fix it, just wanted to have an issue so I didn't forget.

By the way, what is the tyOpt type that can be found next to seq in the code?

[Araq]

By the way, what is the tyOpt type that can be found next to seq in the code?

It is a builtin "opt[T]" type, it's dead code.

[PMunch]

Right, should I just make that into an assert false or something then?