distinction between re.replace and re.replacef; speed improvements for re module

This commit is contained in:
Araq
2011-03-06 21:16:05 +01:00
parent e424e13bd9
commit 3005955d20
23 changed files with 206 additions and 171 deletions

View File

@@ -1,7 +1,7 @@
#
#
# Nimrod's Runtime Library
# (c) Copyright 2010 Andreas Rumpf
# (c) Copyright 2011 Andreas Rumpf
#
# See the file "copying.txt", included in this
# distribution, for details about the copyright.
@@ -32,46 +32,57 @@ type
reIgnoreCase = 0, ## do caseless matching
reMultiLine = 1, ## ``^`` and ``$`` match newlines within data
reDotAll = 2, ## ``.`` matches anything including NL
reExtended = 3 ## ignore whitespace and ``#`` comments
reExtended = 3, ## ignore whitespace and ``#`` comments
reStudy = 4 ## study the expression (may be omitted if the
## expression will be used only once)
TRegExDesc {.pure, final.} = object
h: PPcre
e: ptr TExtra
TRegEx* = ref TRegExDesc ## a compiled regular expression
EInvalidRegEx* = object of EInvalidValue
## is raised if the pattern is no valid regular expression.
proc raiseInvalidRegex(msg: string) {.noinline, noreturn.} =
var e: ref EInvalidRegEx
new(e)
e.msg = msg
raise e
proc rawCompile(pattern: string, flags: cint): PPcre =
var
msg: CString
offset: cint
com = pcre.Compile(pattern, flags, addr(msg), addr(offset), nil)
if com == nil:
var e: ref EInvalidRegEx
new(e)
e.msg = $msg & "\n" & pattern & "\n" & repeatChar(offset) & "^\n"
raise e
return com
result = pcre.Compile(pattern, flags, addr(msg), addr(offset), nil)
if result == nil:
raiseInvalidRegEx($msg & "\n" & pattern & "\n" & repeatChar(offset) & "^\n")
proc finalizeRegEx(x: TRegEx) =
# XXX This is a hack, but PCRE does not export its "free" function properly.
# Sigh. The hack relies on PCRE's implementation (see ``pcre_get.c``).
# Fortunately the implementation is unlikely to change.
pcre.free_substring(cast[cstring](x.h))
if not isNil(x.e):
pcre.free_substring(cast[cstring](x.e))
proc re*(s: string, flags = {reExtended}): TRegEx =
proc re*(s: string, flags = {reExtended, reStudy}): TRegEx =
## Constructor of regular expressions. Note that Nimrod's
## extended raw string literals support this syntax ``re"[abc]"`` as
## a short form for ``re(r"[abc]")``.
new(result, finalizeRegEx)
result.h = rawCompile(s, cast[cint](flags))
result.h = rawCompile(s, cast[cint](flags - {reStudy}))
if reStudy in flags:
var msg: cstring
result.e = pcre.study(result.h, 0, msg)
if not isNil(msg): raiseInvalidRegex($msg)
proc matchOrFind(s: string, pattern: TRegEx, matches: var openarray[string],
start, flags: cint): cint =
var
rawMatches: array[0..maxSubpatterns * 3 - 1, cint]
res = pcre.Exec(pattern.h, nil, s, len(s), start, flags,
res = pcre.Exec(pattern.h, pattern.e, s, len(s), start, flags,
cast[ptr cint](addr(rawMatches)), maxSubpatterns * 3)
if res < 0'i32: return res
for i in 1..int(res)-1:
@@ -83,13 +94,13 @@ proc matchOrFind(s: string, pattern: TRegEx, matches: var openarray[string],
proc findBounds*(s: string, pattern: TRegEx, matches: var openarray[string],
start = 0): tuple[first, last: int] =
## returns the starting position and end position of ``pattern`` in ``s``
## returns the starting position and end position of `pattern` in `s`
## and the captured
## substrings in the array ``matches``. If it does not match, nothing
## is written into ``matches`` and (-1,0) is returned.
## substrings in the array `matches`. If it does not match, nothing
## is written into `matches` and ``(-1,0)`` is returned.
var
rawMatches: array[0..maxSubpatterns * 3 - 1, cint]
res = pcre.Exec(pattern.h, nil, s, len(s), start, 0'i32,
res = pcre.Exec(pattern.h, pattern.e, s, len(s), start, 0'i32,
cast[ptr cint](addr(rawMatches)), maxSubpatterns * 3)
if res < 0'i32: return (-1, 0)
for i in 1..int(res)-1:
@@ -98,10 +109,40 @@ proc findBounds*(s: string, pattern: TRegEx, matches: var openarray[string],
if a >= 0'i32: matches[i-1] = copy(s, int(a), int(b)-1)
else: matches[i-1] = ""
return (rawMatches[0].int, rawMatches[1].int - 1)
proc findBounds*(s: string, pattern: TRegEx,
matches: var openarray[tuple[first, last: int]],
start = 0): tuple[first, last: int] =
## returns the starting position and end position of ``pattern`` in ``s``
## and the captured substrings in the array `matches`.
## If it does not match, nothing is written into `matches` and
## ``(-1,0)`` is returned.
var
rawMatches: array[0..maxSubpatterns * 3 - 1, cint]
res = pcre.Exec(pattern.h, pattern.e, s, len(s), start, 0'i32,
cast[ptr cint](addr(rawMatches)), maxSubpatterns * 3)
if res < 0'i32: return (-1, 0)
for i in 1..int(res)-1:
var a = rawMatches[i * 2]
var b = rawMatches[i * 2 + 1]
if a >= 0'i32: matches[i-1] = (int(a), int(b)-1)
else: matches[i-1] = (-1,0)
return (rawMatches[0].int, rawMatches[1].int - 1)
proc findBounds*(s: string, pattern: TRegEx,
start = 0): tuple[first, last: int] =
## returns the starting position of `pattern` in `s`. If it does not
## match, ``(-1,0)`` is returned.
var
rawMatches: array[0..3 - 1, cint]
res = pcre.Exec(pattern.h, nil, s, len(s), start, 0'i32,
cast[ptr cint](addr(rawMatches)), 3)
if res < 0'i32: return (int(res), 0)
return (int(rawMatches[0]), int(rawMatches[1]-1))
proc matchOrFind(s: string, pattern: TRegEx, start, flags: cint): cint =
var rawMatches: array [0..maxSubpatterns * 3 - 1, cint]
result = pcre.Exec(pattern.h, nil, s, len(s), start, flags,
result = pcre.Exec(pattern.h, pattern.e, s, len(s), start, flags,
cast[ptr cint](addr(rawMatches)), maxSubpatterns * 3)
if result >= 0'i32:
result = rawMatches[1] - rawMatches[0]
@@ -139,7 +180,7 @@ proc find*(s: string, pattern: TRegEx, matches: var openarray[string],
## is written into ``matches`` and -1 is returned.
var
rawMatches: array[0..maxSubpatterns * 3 - 1, cint]
res = pcre.Exec(pattern.h, nil, s, len(s), start, 0'i32,
res = pcre.Exec(pattern.h, pattern.e, s, len(s), start, 0'i32,
cast[ptr cint](addr(rawMatches)), maxSubpatterns * 3)
if res < 0'i32: return res
for i in 1..int(res)-1:
@@ -219,31 +260,64 @@ proc endsWith*(s: string, suffix: TRegEx): bool =
for i in 0 .. s.len-1:
if matchLen(s, suffix, i) == s.len - i: return true
proc replace*(s: string, sub: TRegEx, by: string): string =
## Replaces `sub` in `s` by the string `by`. Captures can be accessed in `by`
## with the notation ``$i`` and ``$#`` (see strutils.`%`). Examples:
proc replace*(s: string, sub: TRegEx, by = ""): string =
## Replaces `sub` in `s` by the string `by`. Captures cannot be
## accessed in `by`. Examples:
##
## .. code-block:: nimrod
## "var1=key; var2=key2".replace(re"(\w+)'='(\w+)", "$1<-$2$2")
## "var1=key; var2=key2".replace(re"(\w+)'='(\w+)")
##
## Results in:
##
## .. code-block:: nimrod
##
## "var1<-keykey; val2<-key2key2"
## "; "
result = ""
var prev = 0
while true:
var match = findBounds(s, sub, prev)
if match.first < 0: break
add(result, copy(s, prev, match.first-1))
add(result, by)
prev = match.last + 1
add(result, copy(s, prev))
proc replacef*(s: string, sub: TRegEx, by: string): string =
## Replaces `sub` in `s` by the string `by`. Captures can be accessed in `by`
## with the notation ``$i`` and ``$#`` (see strutils.`%`). Examples:
##
## .. code-block:: nimrod
## "var1=key; var2=key2".replace(re"(\w+)'='(\w+)", "$1<-$2$2")
##
## Results in:
##
## .. code-block:: nimrod
##
## "var1<-keykey; val2<-key2key2"
result = ""
var i = 0
var caps: array[0..maxSubpatterns-1, string]
while i < s.len:
var x = matchLen(s, sub, caps, i)
if x <= 0:
add(result, s[i])
inc(i)
else:
addf(result, by, caps)
inc(i, x)
# copy the rest:
add(result, copy(s, i))
var prev = 0
while true:
var match = findBounds(s, sub, caps, prev)
if match.first < 0: break
add(result, copy(s, prev, match.first-1))
addf(result, by, caps)
prev = match.last + 1
add(result, copy(s, prev))
when false:
result = ""
var i = 0
var caps: array[0..maxSubpatterns-1, string]
while i < s.len:
var x = matchLen(s, sub, caps, i)
if x <= 0:
add(result, s[i])
inc(i)
else:
addf(result, by, caps)
inc(i, x)
# copy the rest:
add(result, copy(s, i))
proc parallelReplace*(s: string, subs: openArray[
tuple[pattern: TRegEx, repl: string]]): string =
@@ -376,8 +450,10 @@ when isMainModule:
assert false
assert "var1=key; var2=key2".endsWith(re"\w+=\w+")
assert("var1=key; var2=key2".replace(re"(\w+)=(\w+)", "$1<-$2$2") ==
assert("var1=key; var2=key2".replacef(re"(\w+)=(\w+)", "$1<-$2$2") ==
"var1<-keykey; var2<-key2key2")
assert("var1=key; var2=key2".replace(re"(\w+)=(\w+)", "$1<-$2$2") ==
"$1<-$2$2; $1<-$2$2")
for word in split("00232this02939is39an22example111", re"\d+"):
writeln(stdout, word)

View File

@@ -37,7 +37,7 @@
when defined(WIN32):
const dllName = "tcl(85|84|83|82|81|80).dll"
elif defined(macosx):
const dllName = "libtcl(8.5|8.4|8.3|8.2|8.1).dynlib"
const dllName = "libtcl(8.5|8.4|8.3|8.2|8.1).dylib"
else:
const dllName = "libtcl(8.5|8.4|8.3|8.2|8.1).so.(1|0)"

View File

@@ -1,7 +1,7 @@
#
#
# Nimrod's Runtime Library
# (c) Copyright 2010 Andreas Rumpf
# (c) Copyright 2011 Andreas Rumpf
#
# See the file "copying.txt", included in this
# distribution, for details about the copyright.

View File

@@ -874,7 +874,7 @@ proc endsWith*(s: string, suffix: TPeg, start = 0): bool {.
for i in start .. s.len-1:
if matchLen(s, suffix, i) == s.len - i: return true
proc replace*(s: string, sub: TPeg, by: string): string {.
proc replacef*(s: string, sub: TPeg, by: string): string {.
nosideEffect, rtl, extern: "npegs$1".} =
## Replaces `sub` in `s` by the string `by`. Captures can be accessed in `by`
## with the notation ``$i`` and ``$#`` (see strutils.`%`). Examples:
@@ -898,7 +898,23 @@ proc replace*(s: string, sub: TPeg, by: string): string {.
else:
addf(result, by, caps)
inc(i, x)
# copy the rest:
add(result, copy(s, i))
proc replace*(s: string, sub: TPeg, by = ""): string {.
nosideEffect, rtl, extern: "npegs$1".} =
## Replaces `sub` in `s` by the string `by`. Captures cannot be accessed
## in `by`.
result = ""
var i = 0
var caps: array[0..maxSubpatterns-1, string]
while i < s.len:
var x = matchLen(s, sub, caps, i)
if x <= 0:
add(result, s[i])
inc(i)
else:
addf(result, by, caps)
inc(i, x)
add(result, copy(s, i))
proc parallelReplace*(s: string, subs: openArray[
@@ -1691,7 +1707,7 @@ when isMainModule:
"""
assert($g2 == "((A B) / (C D))")
assert match("cccccdddddd", g2)
assert("var1=key; var2=key2".replace(peg"{\ident}'='{\ident}", "$1<-$2$2") ==
assert("var1=key; var2=key2".replacef(peg"{\ident}'='{\ident}", "$1<-$2$2") ==
"var1<-keykey; var2<-key2key2")
assert "var1=key; var2=key2".endsWith(peg"{\ident}'='{\ident}")
@@ -1722,7 +1738,7 @@ when isMainModule:
assert match("EINE ÜBERSICHT UND AUSSERDEM", peg"(\upper \white*)+")
assert(not match("456678", peg"(\letter)+"))
assert("var1 = key; var2 = key2".replace(
assert("var1 = key; var2 = key2".replacef(
peg"\skip(\s*) {\ident}'='{\ident}", "$1<-$2$2") ==
"var1<-keykey;var2<-key2key2")

View File

@@ -687,7 +687,7 @@ proc contains*(s: string, chars: set[char]): bool {.noSideEffect.} =
## Same as ``find(s, chars) >= 0``.
return find(s, chars) >= 0
proc replace*(s, sub, by: string): string {.noSideEffect,
proc replace*(s, sub: string, by = ""): string {.noSideEffect,
rtl, extern: "nsuReplaceStr".} =
## Replaces `sub` in `s` by the string `by`.
var a: TSkipTable
@@ -800,7 +800,7 @@ proc escape*(s: string, prefix = "\"", suffix = "\""): string {.noSideEffect,
## by ``\xHH`` where ``HH`` is its hexadecimal value.
## The procedure has been designed so that its output is usable for many
## different common syntaxes. The resulting string is prefixed with
## ``prefix`` and suffixed with ``suffix``. Both may be empty strings.
## `prefix` and suffixed with `suffix`. Both may be empty strings.
result = prefix
for c in items(s):
case c

View File

@@ -65,11 +65,11 @@ proc `not` *(x: bool): bool {.magic: "Not", noSideEffect.}
proc `and`*(x, y: bool): bool {.magic: "And", noSideEffect.}
## Boolean ``and``; returns true iff ``x == y == true``.
## Evaluation is short-circuited: this means that if ``x`` is false,
## Evaluation is lazy: if ``x`` is false,
## ``y`` will not even be evaluated.
proc `or`*(x, y: bool): bool {.magic: "Or", noSideEffect.}
## Boolean ``or``; returns true iff ``not (not x and not y)``.
## Evaluation is short-circuited: this means that if ``x`` is true,
## Evaluation is lazy: if ``x`` is true,
## ``y`` will not even be evaluated.
proc `xor`*(x, y: bool): bool {.magic: "Xor", noSideEffect.}
## Boolean `exclusive or`; returns true iff ``x != y``.
@@ -623,7 +623,7 @@ template `not_in` * (x, y: expr): expr = not contains(y, x)
proc `is` *[T, S](x: T, y: S): bool {.magic: "Is", noSideEffect.}
template `is_not` *(x, y: expr): expr = not (x is y)
proc cmp*[T, S: typeDesc](x: T, y: S): int {.procvar.} =
proc cmp*[T](x, y: T): int {.procvar.} =
## Generic compare proc. Returns a value < 0 iff x < y, a value > 0 iff x > y
## and 0 iff x == y. This is useful for writing generic algorithms without
## performance loss. This generic implementation uses the `==` and `<`
@@ -1034,9 +1034,6 @@ iterator countup*[S, T](a: S, b: T, step = 1): T {.inline.} =
while res <= b:
yield res
inc(res, step)
# we cannot use ``for x in a..b: `` here, because that is not
# known in the System module
proc min*(x, y: int): int {.magic: "MinI", noSideEffect.}
proc min*(x, y: int8): int8 {.magic: "MinI", noSideEffect.}

View File

@@ -31,7 +31,7 @@ elif defined(useNimRtl):
when hostOS == "windows":
const nimrtl* = "nimrtl.dll"
elif hostOS == "macosx":
const nimrtl* = "nimrtl.dynlib"
const nimrtl* = "nimrtl.dylib"
else:
const nimrtl* = "libnimrtl.so"
{.pragma: rtl, importc: "nimrtl_$1", dynlib: nimrtl.}

View File

@@ -27,7 +27,7 @@ when defined(windows):
clarodll = "claro.dll"
elif defined(macosx):
const
clarodll = "libclaro.dynlib"
clarodll = "libclaro.dylib"
else:
const
clarodll = "libclaro.so"

View File

@@ -39,7 +39,7 @@ when not defined(pcreDll):
when hostOS == "windows":
const pcreDll = "pcre3.dll"
elif hostOS == "macosx":
const pcreDll = "libpcre.dynlib"
const pcreDll = "libpcre(.3|).dylib"
else:
const pcreDll = "libpcre.so(.3|)"
@@ -275,7 +275,7 @@ proc maketables*(): ptr char{.cdecl, importc: "pcre_maketables",
dynlib: pcredll.}
proc refcount*(a2: ptr TPcre, a3: cint): cint{.cdecl, importc: "pcre_refcount",
dynlib: pcredll.}
proc study*(a2: ptr TPcre, a3: cint, a4: cstringArray): ptr Textra{.cdecl,
proc study*(a2: ptr TPcre, a3: cint, a4: var cstring): ptr Textra{.cdecl,
importc: "pcre_study", dynlib: pcredll.}
proc version*(): cstring{.cdecl, importc: "pcre_version", dynlib: pcredll.}

View File

@@ -39,7 +39,7 @@ when defined(WIN32):
dllName = "tcl(85|84|83|82|81|80).dll"
elif defined(macosx):
const
dllName = "libtcl(8.5|8.4|8.3|8.2|8.1).dynlib"
dllName = "libtcl(8.5|8.4|8.3|8.2|8.1).dylib"
else:
const
dllName = "libtcl(8.5|8.4|8.3|8.2|8.1).so.(1|0)"

View File

@@ -10,7 +10,7 @@ when not defined(treDll):
when hostOS == "windows":
const treDll = "tre.dll"
elif hostOS == "macosx":
const treDll = "libtre.dynlib"
const treDll = "libtre.dylib"
else:
const treDll = "libtre.so(.5|)"