Mirror of forum.nim-lang.org

7658 :: Regex not hashable?

• lhupfeldt (orginal) [2021-03-18T21:40:13+01:00] view original

  
  import sets
  import re
  
  discard toHashSet([re"a"])
  Run

  .../nim-1.4.4/lib/pure/collections/hashcommon.nim(69, 12) Error: type mismatch: got <Regex>
  but expected one of:
  proc hash(sBuf: string; sPos, ePos: int): Hash
    first type mismatch at position: 1
  ...

---

• cblake (orginal) [2021-03-18T22:38:41+01:00] view original

  Looks like there is no proc hash(x: Regex) defined..Probably no == either. You need those and note that it may not be trivial if you want equivalent expressions to report as equal. Then you need to concoct a proc hash honoring that.

  You might think "oh, I just want the string used to build the regex", but that is not saved in a Regex and so may be long gone. You may be able to just use the string spelling of the original expression (after the re), not the compiled Regex , but it's hard to say without more context.

---

• lhupfeldt (orginal) [2021-03-18T23:31:11+01:00] view original

  Thank you for the reply. Is there no Hashable interface in nim? It is quite a surprise to me that a seemingly simple thing like putting a Regex in a HashSet does not work out of the box. I'm also surprised that the pattern string is not stored with the Regex, I would think that it would be needed for string representaion of the Regex. Obviously there would be an overhead there which may be unwanted in some applications. I'm only using the HashSet for uniqueness, and i have the string available so I have an easy, but not so nice workaround of using a HashTable instead.

---

• cblake (orginal) [2021-03-18T23:39:43+01:00] view original

  The "interface" is really just in-scope proc hash and proc == overloads for the type in question. proc $ is also undefined:

  import re
  let x = re"hi"
  # echo x # fails, no `$`
  echo x.repr # a hint why above fails :)
  Run

  Sounds like you have a solution, though. Good luck!

---

• lhupfeldt (orginal) [2021-03-19T00:07:27+01:00] view original

  But the problem with overloading hash`and `== is that there seems to be nothing to hash on. If the original pattern string was there it would be simple.

---

• xigoi (orginal) [2021-03-19T00:23:25+01:00] view original

  If you want regexes to save the original string, you can do it:

  
  type RegexWithOriginalString = (Regex, string)
  Run

---