Mirror of forum.nim-lang.org

2327 :: Mystic strings

• Stefan_Salewski (orginal) [2016-06-18T21:21:07+02:00] view original

  Seems that I really know not much about Nim strings...

  var FirmwareTableProviderSignature: uint32
  var a, b, c, d: uint32
  FirmwareTableProviderSignature = cast[uint32]("ACPI".cstring)
  
  echo 'A'.ord shl 0 + 'C'.ord shl 8 + 'P'.ord shl 16 + 'I'.ord shl 24
  
  a = cast[uint32]("ACPI")
  b = cast[uint32]("ACPI".cstring)
  c = cast[uint32]("ACPI"[0])
  d = cast[uint32]("ACPI"[0 .. 3])
  
  echo a
  echo b
  echo c
  echo d
  Run

  $ ./t
  1229996865
  4323872
  4323921
  65
  1255772240

  The last value is arbitrary. Will think or read about it tomorrow -- indeed I was going to try castuint32 which does not compile.

---

• Varriount (orginal) [2016-06-18T23:11:02+02:00] view original

  The assignment to variable "d" creates a new string object and copies characters 0-3 from the original string into it. Casting the string to a uint32 causes the pointer to the string information to be viewed as an unsigned integer, hence the arbitrary value.

---

• OderWat (orginal) [2016-06-18T23:24:13+02:00] view original

  echo 'A'.ord shl 0 + 'C'.ord shl 8 + 'P'.ord shl 16 + 'I'.ord shl 24 var a,b,c,d: uint32 let x = "ACPI" a = cast[ptr uint32](x[0].unsafeAddr)[] b = cast[ptr uint32](x.cstring)[] c = cast[ptr uint32]("ACPI".cstring)[] d = cast[ptr uint32](cast[int]("ACPI".cstring))[] echo a echo b echo c # 65 ??? echo d

  Run

---

• Varriount (orginal) [2016-06-19T00:06:56+02:00] view original

  "A" is represented by the unsigned integer 65 in utf-8. On that particular line you are covering a string to a cstring, converting that into a pointer to a unsigned integer, then dereferencing the pointer. Since a cstring is a pointer to a character array, dereferencing the pointer will give you the first character.

  There is a possible bug this program reveals though - the type of c looks to be a uint8, when it should be something else. This is because the value printed should be a amalgamation of the first four character's integer form (8-bit characters times four equals 32).

---

• OderWat (orginal) [2016-06-19T18:22:00+02:00] view original

  @variount yes that possible bug is why I wrote "65 ???" ... imho it should be the same result as in d.

  Something similar happens with:

  var a = cast[ptr uint32](cast[int](['A','C','P','I']))[]
  echo a
  Run

  Which crashes without the intermediate cast to int.

  Note that casting the intermediate state to "pointer" or any "ptr type" does not work either.

  The cast to ptr uint32 of the c string generates.

  x = *"ACPI"
  Run

  While the other version casts the cstring pointer to an int and this than to the uint32 pointer.

  x = *((NU32*) ((NI) ("ACPI")));
  Run

  It seems like the Nim compiler is missing that the cstring is actually a pointer to a char array and as such needs a cast to the uint32 type to work correctly. I don't know if it is supposed to know that "cstring" is a "ptr char array" in the context of casts.

---