Updated 2011-10-23 11:17:13 by RLE

Richard Suchenwirth 2002-07-17 - Since Tcl 8.4, lindex and the new lset command take multiple indices to retrieve or modify an element in a nested list, which might be used to represent a multi-dimensional table (matrix, database) or a tree structure. To complement these two new goodies, here's a recursive lsearch that returns an index list, of the first occurrence of the wanted item, as can be used by lindex or lset, or -1 if the element was not found.

Note that the search goes down as long as an element can be interpreted as a list of more than one element, so intended strings like "John F Kennedy" will also be searched, e.g. for F. Note also that the result often is a list, so comparisons should be limited to == -1 or != -1.
 proc lrsearch {list el {prefix {}} } {
    set pos [lsearch $list $el]
    if {$pos != -1} {return [concat $prefix $pos]}
    for {set i 0} {$i<[llength $list]} {incr i} {
        set ilist [lindex $list $i]
        if {![atomic? $ilist]} {
            set pos [lrsearch $ilist $el [concat $prefix $i]]
            if {$pos != -1} {return $pos}
        }
    }
    return -1
 }

To prevent endless recursion, but still handle lists that have only one element on higher level, the atomic? test was introduced: it is true if the first element of the list is equal to the list itself, so that no further lindexing could bring any new facts.
 proc atomic? {list} {string equal $list [lindex $list 0]}

 % lrsearch {a b {c {d e}}} a
 0
 % lrsearch {a b {c {d e}}} b
 1
 % lrsearch {a b {c {d e}}} c
 2 0
 % lrsearch {a b {c {d e}}} d
 2 1 0
 % lrsearch {a b {c {d e}}} e
 2 1 1
 % lrsearch {a b {c {d e}}} f
 -1
 % lrsearch {a {{b c}} d} b
 1 0 0, e.g. set s {a b {a b {a b c {a, e.g. set s {a b {a b {a b c {a d e c {a b c} c}}} d e} d e c {a b c} c}}} d e}

Michael Schlenker likes to do things a bit faster; this is between 5 and 25 percent faster (with 8.3 and 8.4b1) than the solution above, by using foreach/incr instead of for/lindex:
 proc lrsearch2 {list el {prefix {}} } {
    set pos [lsearch $list $el]
    if {-1 != $pos} {return [concat $prefix $pos]}
    set i 0
    foreach ilist $list {
        if {![atomic? $ilist]} {
            set pos [lrsearch2 $ilist $el [concat $prefix $i]]
            if {-1 != $pos} {return $pos}
        }
        incr i
    }
    return -1gs a bit faster; this is between 5 and 25 percent faster (with 8.3 and 8.4b1)
than the solution above, by using [f
 }

[serol] 2010-11-19: I needed a slightly modified lrsearch to get the number of occurrences of string in a nested list.
 proc lrsearch2 {liste el} {        
  #get occurences in liste        
  set count [llength [lsearch -all -inline $liste $el]]

  foreach ilist $liste {
   if {![atomic? $ilist]} {
    incr count [lrsearch2 $ilist $el]                  
   }
  }
  return $count
 }