Hi,

I knocked up some code to extract the plain text from an RTF document. It works for my purposes but would like to know if anyone can find a case where it does a bad job.

Copy your .rtf file into a flat array, eg x(1)=first line, x(2)=second line then:

d ##class(yourclass).StripRTF(.x,.y)

and you'll get the plain text in y

/// accepts an RTF doc in array form, doesn't care if lines are split across array members
/// returns plain text from the doc with one array item per line 
/// quit value is number of lines in array
ClassMethod StripRTF(ByRef rtfText = "", ByRef %plainText) As %Integer
{
 // use this code to view an rtf doc split into groups and indented when group level goes up
 // S IN=0,C=0 F I=1:1 S C=##class(ZSS.SMTP).EXTRACT(.x,I) Q:C="" I C="{" {S IN=IN+1 W !?IN*2,"(",IN,")"} W C I C="}" {W "(",IN,")" S IN=IN-1 W !?IN*2}
 n %line
 kill %plainText Set %line=0
 // you could add some speed here by working out how to set characters in the R array to null and
 // reducing the i pointer accordingly
 m R=rtfText
 d ..Brace(.R,$i(i))
 q %line
}

/// A bit like $Extract but the first argument is a single level array, like array(1)="some text", array(2)="some more text"
/// ..EXTRACT(.array,9,11) would return "tso" as the 9th, 10th, and 11th characters of the data in the array
ClassMethod EXTRACT(ByRef array, from As %Integer, to As %Integer, SetToNull = 0) As %String
{
 i '$d(to) s to=from
 S UncleLen=from-to+1 // length of string to return
 s string=""
 f index=1:1 {
  q:'$d(array(index))
  s short=array(index)
  i from'>$l(short) {
   s string=string_$e(short,from,to)
   // mimic SET $EXTRACT
   I SetToNull S $e(array(index),from,to)=""
   q:$l(string)=UncleLen
  }
  s from=from-$l(short),to=to-$l(short)
  i from<0 s from=0
 }
 q string
}
/// discard everything between two braces including the braces
/// done by either moving i to the last brace
/// or if SetToNull is passed as 1 then actually removing the characters from the RTF array
ClassMethod Discard(ByRef rtfText, ByRef i, SetToNull = 0)
{
 s inGroup=1,start=i,stop=i,discarded=""
 f  {
  s discard=..EXTRACT(.rtfText,$i(i))
  q:discard=""
  i discard="}" {
   s inGroup=inGroup-1
  }
  i discard="{" s inGroup=inGroup+1
 q
 }
}
i SetToNull d ..EXTRACT(.rtfText,start,i,1)
}
/// return the contents of a Slash
ClassMethod Slash(ByRef R, ByRef i) As %String
{
 s string="\"
 F  {
  Set char=..EXTRACT(.R, $I(i))
  q:char="" // should never happen but don't want to get stuck in a loop because of a bad file
  i char="}" s i=i-1 q
  i char=" " q
  i char="\" s i=i-1 q
  i char="{" {
   i string?1"\"1a.an {
    // everything inside something like \abc1{this stuff}
    // example {\fonttbl{\fprq2{02020603050405020304}TimesNew Roman;} should all be discarded
    d ..Discard(.R,.i)
    s string=$c(127)
    q
   }
  s char=..Brace(.R,.i)
  }
  i char="*",..EXTRACT(.R, i-2,i+1)="{\*\" {
   // "{\*\" at the start of a brace means the whole brace including nested braces can be ignored
   d ..Discard(.R,.i)
   s string=$c(127)
   q
  }
  i char="'" {
   s string=string_char_..EXTRACT(.R, $I(i),$I(i))
   q
  }
  s string=string_char
  // everything in here should disappear because it looks like {\stylesheet{
  i string="\stylesheet"!(string="\info") {
   d ..Discard(.R,.i)
   s string=$c(127)
   q
  }
 }
 //
 // add stuff here for special characters represented by \codeword
 // you could put them on a global
 //
 i string'="" {
  try {
  // catching stupid subscripts
  s string=$G(^GMAT("RTF","special characters",string),string)
  }
  catch {
   // do nowt
  }
 }
 i string="\lquote" s string="'"
 i string="\rquote" s string="'"
 i string="\ldblquote" s string=""""
 i string="\rdblquote" s string=""""
 i string="\'93" s string=""""  // $c($zh("93")) should be a left double quote
 i string="\'94" s string=""""
 i string?1"\'"2an s string=$c($zh($p(string,"'",2)))
 i string="\par" s string=$c(13,10)
 // REMOVING \codename and \codename1 and \codename1;
 i string ?1"\"1a.an.1";" q ""
 // REMOVING \codename-20
 i string ?1"\"1.a1"-"1.n.1";" q ""
 i string="{}" q ""
 q string
 }
/// return the contents of a pair of braces (troosers!)
ClassMethod Brace(ByRef R, ByRef i) As %String
{
 s string="{"
 F  {
  Set char=..EXTRACT(.R, $I(i),i+1)
  // reached the end and there will be a loose "}"
  i char="",string="}" s string=""
  q:char=""
  // escaped characters that should be allowed through to the text
  s escape=0
  i $lf($lb("\\","\{","\}"),char) s i=i+1,escape=1
  e  s char=$e(char)
  i char="\" {
  s char=..Slash(.R,.i)
  // we hit a ..Discard so remove the brace before it
  i char=$c(127) s char="",$e(string,*)=""
  s string=string_char
  continue
 }
 i char="{" {
  s char=..Brace(.R,.i)
  i char=$c(13,10) {
   //works for 1st line only s %plainText($i(%line))=string,string="",char=""
  f {
   s %plainText($i(%line))=$p(string,$c(13,10),1)
   s string=$p(string,$c(13,10),2,*)
   q:string=""
  }
  s char=""
  }
  s string=string_char
  continue
  }
  // HEX ascii
  i char="\'" {
   s char=$C($ZHEX(..EXTRACT(.R, $I(i),$I(i))))
  }
  s string=string_$e(char,*)
  i string="{}" s string=""
  q:char="}"
 }
 i string="{}" {
  s string=""
 }
 i $e(string)="{",$e(string,*)="}" {
  s string=$e(string,2,*-1)
 }
 q string
}

Hi Julius, I was playing Devil's "avocado 😁" with that. One could argue the case for that being the correct approach. One could also argue for Paul's unit tests based on seeing no reason why rotational symmetry couldn't apply. (I think my solution for that is 225 and doesn't beat my friend Paul's answer so I don't like it.) The point is that the spiral has to get smaller somewhere but we haven't been given a clear enough rule to follow. In the example in the question, the spiral only turns inward when it hits a used letter or a corner - a dead end as @Robert Cemper puts it. The unit tests below could also be valid. (My solution for this is 199)

There's an unwritten rule in the question that could be "as a human, make a decision when to turn inwards to make sense out of the order of letters". How do you code that?

        Set matrix($INCREMENT(matrix)) = "A,B,C"
        Set matrix($INCREMENT(matrix)) = "H,I,D"
        Set matrix($INCREMENT(matrix)) = "G,F,E"
        d $$$AssertEquals(..Solution(.matrix, 1, 1), "ABCDEFGHI")
        d $$$AssertEquals(..Solution(.matrix, 1, 2), "BCDEFGHA") // !!?
        d $$$AssertEquals(..Solution(.matrix, 1, 3), "CDEFGHABI")
        d $$$AssertEquals(..Solution(.matrix, 2, 3), "DEFGHABC") // !!?
        d $$$AssertEquals(..Solution(.matrix, 3, 3), "EFGHABCDI")
        d $$$AssertEquals(..Solution(.matrix, 3, 2), "FGHABCDE") // !!?
        d $$$AssertEquals(..Solution(.matrix, 3, 1), "GHABCDEFI")
        d $$$AssertEquals(..Solution(.matrix, 2, 1), "HABCDEFG") // !!?

You are correct Paul, but there's a problem with other tests too. The one with 3,1 above could return GHABCDI because you could argue that when the spiral leaves an outer edge it should never return to it. If the test that starts at 1,2 doesn't return to row 1 then why should 1,3 return to row 1 after leaving? These could be the correct tests:

        Set matrix($INCREMENT(matrix)) = "A,B,C"
        Set matrix($INCREMENT(matrix)) = "H,I,D"
        Set matrix($INCREMENT(matrix)) = "G,F,E"
        d $$$AssertEquals(..Solution(.matrix, 1, 1), "ABCDEFGHI")
        d $$$AssertEquals(..Solution(.matrix, 1, 2), "BCDEFGHI")
        d $$$AssertEquals(..Solution(.matrix, 1, 3), "CDEFGHI")
        d $$$AssertEquals(..Solution(.matrix, 2, 3), "DEFGHI")
        d $$$AssertEquals(..Solution(.matrix, 3, 3), "EFGHI")
        d $$$AssertEquals(..Solution(.matrix, 3, 2), "FGHI")
        d $$$AssertEquals(..Solution(.matrix, 3, 1), "GHI")
        d $$$AssertEquals(..Solution(.matrix, 2, 1), "HI")