From 784857e9faab18ded1de77b36fd2ada32796390f Mon Sep 17 00:00:00 2001
From: Eli Barzilay <eli@barzilay.org>
Date: Wed, 23 May 2012 18:38:09 -0400
Subject: [PATCH] Add `string-split'.

---
 collects/racket/string.rkt                   | 24 ++++++++++----
 collects/scribblings/reference/strings.scrbl | 34 ++++++++++++++++----
 collects/tests/racket/string.rktl            | 12 +++++--
 3 files changed, 55 insertions(+), 15 deletions(-)

diff --git a/collects/racket/string.rkt b/collects/racket/string.rkt
index 6cae70ddb9..d1b762b038 100644
--- a/collects/racket/string.rkt
+++ b/collects/racket/string.rkt
@@ -1,6 +1,7 @@
 #lang racket/base
 
-(provide string-append* string-join string-trim string-normalize-spaces)
+(provide string-append* string-join string-trim string-normalize-spaces
+         string-split)
 
 (define string-append*
   (case-lambda [(strs) (apply string-append strs)] ; optimize common case
@@ -67,11 +68,20 @@
         [r         (substring str 0 r)]
         [else      str]))
 
+(define (internal-split who str sep trim? +?)
+  (define rxs (get-rxs who sep +?))
+  (define-values [l r]
+    (if trim? (internal-trim who str sep #t #t (cdr rxs)) (values #f #f)))
+  (define strs (regexp-split (car rxs) str (or l 0) r))
+  ;; Seems to make more sense for these functions (eg, this corresponds to
+  ;; simple uses where `string-split' in Emacs uses t for `omit-nulls' (but we
+  ;; don't do that for all nulls).)
+  (if (equal? strs '("")) '() strs))
+
+(define (string-split str [sep none] #:trim? [trim? #t] #:repeat? [+? #f])
+  (internal-split 'string-split str sep trim? +?))
+
 (define (string-normalize-spaces str [sep none] [space " "]
                                  #:trim? [trim? #t] #:repeat? [+? #f])
-  (define rxs (get-rxs 'string-normalize-spaces sep +?))
-  (define-values [l r]
-    (if trim?
-      (internal-trim 'string-normalize-spaces str sep #t #t (cdr rxs))
-      (values #f #f)))
-  (string-join (regexp-split (car rxs) str (or l 0) r) space))
+  (string-join (internal-split 'string-normalize-spaces str sep trim? +?)
+               space))
diff --git a/collects/scribblings/reference/strings.scrbl b/collects/scribblings/reference/strings.scrbl
index fe78cd2755..156e16c886 100644
--- a/collects/scribblings/reference/strings.scrbl
+++ b/collects/scribblings/reference/strings.scrbl
@@ -429,6 +429,29 @@ of matches is trimmed.  (Note that with a regexp separator you can use
   (string-trim "aaaxaayaa" "aa")
 ]}
 
+@defproc[(string-split [str string?]
+                       [sep (or/c string? regexp?) #px"\\s+"]
+                       [#:trim? trim? any/c #t]
+                       [#:repeat? repeat? any/c #f])
+         (listof string?)]{
+
+Splits the input @racket[str] on whitespaces, returning a list of
+strings.  The input is trimmed first.
+
+Similarly to @racket[string-trim], @racket[sep] can be given as a string
+or a (p)regexp to use a different separator, and @racket[repeat?]
+controls matching repeated sequences.  @racket[trim?] determines whether
+trimming is done (the default).
+
+@mz-examples[#:eval string-eval
+  (string-split "  foo bar  baz \r\n\t")
+  (string-split "  ")
+  (string-split "  " #:trim? #f)
+]
+
+(Note that unlike @racket[regexp-split], an empty input string results
+in an empty list.)}
+
 @defproc[(string-normalize-spaces [str string?]
                                   [sep (or/c string? regexp?) #px"\\s+"]
                                   [space string? " "]
@@ -440,15 +463,14 @@ Normalizes spaces in the input @racket[str] by trimming it (using
 @racket[string-trim]) and replacing all whitespace sequences in the
 result with a single space.
 
-Similarly to @racket[string-trim], @racket[sep] can be given as a string
-or a (p)regexp, and @racket[repeat?] controls matching repeated
-sequences.  In addition, you can specify @racket[space] for an alternate
-space replacement.  @racket[trim?] determines whether trimming is done
-(the default).
+You can specify @racket[space] for an alternate space replacement.
 
 @mz-examples[#:eval string-eval
   (string-normalize-spaces "  foo bar  baz \r\n\t")
-]}
+]
+
+Note that this is the same as
+@racket[(string-join (string-split str sep ....) space)]}
 
 
 @close-eval[string-eval]
diff --git a/collects/tests/racket/string.rktl b/collects/tests/racket/string.rktl
index 1a57a2110d..cc79ebc569 100644
--- a/collects/tests/racket/string.rktl
+++ b/collects/tests/racket/string.rktl
@@ -430,7 +430,15 @@
   ;; this should return "" or "ba" (could also be "ab"), but it seems sensible
   ;; to do this (I haven't seen any existing trimmers that make any relevant
   ;; decision on this)
-  (test "" string-trim "ababa" "aba")
-  )
+  (test "" string-trim "ababa" "aba"))
+
+;; ---------- string-split ----------
+(let ()
+  (for ([s (in-list '("x y z" " x y z "  "\nx y z" "  \t x\r\r\ry    z\n"))])
+    (test '("x" "y" "z") string-split s))
+  (for ([s (in-list '(" " "   "  "\n\t\r"))])
+    (test '() string-split s))
+  (test '("x" "y" "z") string-split "axayaza" "a")
+  (test '("" "x" "y" "z" "") string-split "axayaza" "a" #:trim? #f))
 
 (report-errs)