jyapayne/Nim
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

refactor cmpIgnoreStyle and cmpIgnoreCase (#16399)

Browse source 
* init

* support strutils

* more

* better

* Call len once per string/cstring

* Change var to let

* Compare ternary on first char

* More appropriate param name

* fix

* better

* one test

* impl

* more efficient

* minor

Co-authored-by: Clyybber <darkmine956@gmail.com>
This commit is contained in:
flywind 2020-12-31 04:54:40 -06:00 • committed by GitHub
commit 5fb56a3b2c
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
8 changed files with 141 additions and 147 deletions
Download patch file Download diff file Expand all files Collapse all files

90
lib/pure/cstrutils.nim
View file

@ -12,12 +12,8 @@
## save allocations.
include "system/inclrtl"
import std/private/strimpl
proc toLowerAscii(c: char): char {.inline.} =
if c in {'A'..'Z'}:
result = chr(ord(c) + (ord('a') - ord('A')))
else:
result = c
when defined(js):
proc startsWith*(s, prefix: cstring): bool {.noSideEffect,
@ -25,7 +21,13 @@ when defined(js):
proc endsWith*(s, suffix: cstring): bool {.noSideEffect,
importjs: "#.endsWith(#)".}
proc cmpIgnoreStyle*(a, b: cstring): int {.noSideEffect.} =
cmpIgnoreStyleImpl(a, b)
proc cmpIgnoreCase*(a, b: cstring): int {.noSideEffect.} =
cmpIgnoreCaseImpl(a, b)
# JS string has more operations that might warrant its own module:
# https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/String
else:
@ -57,45 +59,39 @@ else:
inc(i)
if suffix[i] == '\0': return true
proc cmpIgnoreStyle*(a, b: cstring): int {.noSideEffect,
rtl, extern: "csuCmpIgnoreStyle".} =
## Semantically the same as ``cmp(normalize($a), normalize($b))``. It
## is just optimized to not allocate temporary strings. This should
## NOT be used to compare Nim identifier names. use `macros.eqIdent`
## for that. Returns:
##
## | 0 if a == b
## | < 0 if a < b
## | > 0 if a > b
##
## Not supported for JS backend, use `strutils.cmpIgnoreStyle
## <strutils.html#cmpIgnoreStyle%2Cstring%2Cstring>`_ instead.
var i = 0
var j = 0
while true:
while a[i] == '_': inc(i)
while b[j] == '_': inc(j) # BUGFIX: typo
var aa = toLowerAscii(a[i])
var bb = toLowerAscii(b[j])
result = ord(aa) - ord(bb)
if result != 0 or aa == '\0': break
inc(i)
inc(j)
proc cmpIgnoreStyle*(a, b: cstring): int {.noSideEffect,
rtl, extern: "csuCmpIgnoreStyle".} =
## Semantically the same as ``cmp(normalize($a), normalize($b))``. It
## is just optimized to not allocate temporary strings. This should
## NOT be used to compare Nim identifier names. use `macros.eqIdent`
## for that. Returns:
##
## | 0 if a == b
## | < 0 if a < b
## | > 0 if a > b
var i = 0
var j = 0
while true:
while a[i] == '_': inc(i)
while b[j] == '_': inc(j) # BUGFIX: typo
var aa = toLowerAscii(a[i])
var bb = toLowerAscii(b[j])
result = ord(aa) - ord(bb)
if result != 0 or aa == '\0': break
inc(i)
inc(j)
proc cmpIgnoreCase*(a, b: cstring): int {.noSideEffect,
rtl, extern: "csuCmpIgnoreCase".} =
## Compares two strings in a case insensitive manner. Returns:
##
## | 0 if a == b
## | < 0 if a < b
## | > 0 if a > b
##
## Not supported for JS backend, use `strutils.cmpIgnoreCase
## <strutils.html#cmpIgnoreCase%2Cstring%2Cstring>`_ instead.
var i = 0
while true:
var aa = toLowerAscii(a[i])
var bb = toLowerAscii(b[i])
result = ord(aa) - ord(bb)
if result != 0 or aa == '\0': break
inc(i)
proc cmpIgnoreCase*(a, b: cstring): int {.noSideEffect,
rtl, extern: "csuCmpIgnoreCase".} =
## Compares two strings in a case insensitive manner. Returns:
##
## | 0 if a == b
## | < 0 if a < b
## | > 0 if a > b
var i = 0
while true:
var aa = toLowerAscii(a[i])
var bb = toLowerAscii(b[i])
result = ord(aa) - ord(bb)
if result != 0 or aa == '\0': break
inc(i)

30
lib/pure/strutils.nim
View file

@ -81,6 +81,8 @@ when defined(nimVmExportFixed):
include "system/inclrtl"
import std/private/since
from std/private/strimpl import cmpIgnoreStyleImpl, cmpIgnoreCaseImpl
const
Whitespace* = {' ', '\t', '\v', '\r', '\l', '\f'}
@ -319,13 +321,7 @@ func cmpIgnoreCase*(a, b: string): int {.rtl, extern: "nsuCmpIgnoreCase".} =
doAssert cmpIgnoreCase("FooBar", "foobar") == 0
doAssert cmpIgnoreCase("bar", "Foo") < 0
doAssert cmpIgnoreCase("Foo5", "foo4") > 0
var i = 0
var m = min(a.len, b.len)
while i < m:
result = ord(toLowerAscii(a[i])) - ord(toLowerAscii(b[i]))
if result != 0: return
inc(i)
result = a.len - b.len
cmpIgnoreCaseImpl(a, b)
{.push checks: off, line_trace: off.} # this is a hot-spot in the compiler!
# thus we compile without checks here
@ -344,25 +340,7 @@ func cmpIgnoreStyle*(a, b: string): int {.rtl, extern: "nsuCmpIgnoreStyle".} =
runnableExamples:
doAssert cmpIgnoreStyle("foo_bar", "FooBar") == 0
doAssert cmpIgnoreStyle("foo_bar_5", "FooBar4") > 0
var i = 0
var j = 0
while true:
while i < a.len and a[i] == '_': inc i
while j < b.len and b[j] == '_': inc j
var aa = if i < a.len: toLowerAscii(a[i]) else: '\0'
var bb = if j < b.len: toLowerAscii(b[j]) else: '\0'
result = ord(aa) - ord(bb)
if result != 0: return result
# the characters are identical:
if i >= a.len:
# both cursors at the end:
if j >= b.len: return 0
# not yet at the end of 'b':
return -1
elif j >= b.len:
return 1
inc i
inc j
cmpIgnoreStyleImpl(a, b)
{.pop.}
# --------- Private templates for different split separators -----------