#
#
# Nim's Runtime Library
# (c) Copyright 2010 Andreas Rumpf
#
# See the file "copying.txt", included in this
# distribution, for details about the copyright.
#
## This module contains support for a `rope`:idx: data type.
## Ropes can represent very long strings efficiently; especially concatenation
## is done in O(1) instead of O(n). They are essentially concatenation
## trees that are only flattened when converting to a native Nim
## string. The empty string is represented by ``nil``. Ropes are immutable and
## subtrees can be shared without copying.
## Leaves can be cached for better memory efficiency at the cost of
## runtime efficiency.
include "system/inclrtl"
import streams
{.push debugger: off.} # the user does not want to trace a part
# of the standard library!
const
countCacheMisses = false
var
cacheEnabled = false
type
Rope* = ref RopeObj ## empty rope is represented by nil
RopeObj {.acyclic.} = object
left, right: Rope
length: int
data: string # != nil if a leaf
# Note that the left and right pointers are not needed for leafs.
# Leaves have relatively high memory overhead (~30 bytes on a 32
# bit machine) and we produce many of them. This is why we cache and
# share leafs across different rope trees.
# To cache them they are inserted in another tree, a splay tree for best
# performance. But for the caching tree we use the leaf's left and right
# pointers.
proc len*(a: Rope): int {.rtl, extern: "nro$1".} =
## the rope's length
if a == nil: result = 0
else: result = a.length
proc newRope(): Rope = new(result)
proc newRope(data: string): Rope =
new(result)
result.length = len(data)
result.data = data
var
cache {.threadvar.}: Rope # the root of the cache tree
N {.threadvar.}: Rope # dummy rope needed for splay algorithm
when countCacheMisses:
var misses, hits: int
proc splay(s: string, tree: Rope, cmpres: var int): Rope =
var c: int
var t = tree
N.left = nil
N.right = nil # reset to nil
var le = N
var r = N
while true:
c = cmp(s, t.data)
if c < 0:
if (t.left != nil) and (s < t.left.data):
var y = t.left
t.left = y.right
y.right = t
t = y
if t.left == nil: break
r.left = t
r = t
t = t.left
elif c > 0:
if (t.right != nil) and (s > t.right.data):
var y = t.right
t.right = y.left
y.left = t
t = y
if t.right == nil: break
le.right = t
le = t
t = t.right
else:
break
cmpres = c
le.right = t.left
r.left = t.right
t.left = N.right
t.right = N.left
result = t
proc insertInCache(s: string, tree: Rope): Rope =
var t = tree
if t == nil:
result = newRope(s)
when countCacheMisses: inc(misses)
return
var cmp: int
t = splay(s, t, cmp)
if cmp == 0:
# We get here if it's already in the Tree
# Don't add it again
result = t
when countCacheMisses: inc(hits)
else:
when countCacheMisses: inc(misses)
result = newRope(s)
if cmp < 0:
result.left = t.left
result.right = t
t.left = nil
else:
# i > t.item:
result.right = t.right
result.left = t
t.right = nil
proc rope*(s: string = ""): Rope {.rtl, extern: "nro$1Str".} =
## Converts a string to a rope.
if s.len == 0:
result = nil
else:
when nimvm:
# No caching in VM context
result = newRope(s)
else:
if cacheEnabled:
result = insertInCache(s, cache)
cache = result
else:
result = newRope(s)
proc rope*(i: BiggestInt): Rope {.rtl, extern: "nro$1BiggestInt".} =
## Converts an int to a rope.
result = rope($i)
proc rope*(f: BiggestFloat): Rope {.rtl, extern: "nro$1BiggestFloat".} =
## Converts a float to a rope.
result = rope($f)
proc enableCache*() {.rtl, extern: "nro$1".} =
## Enables the caching of leaves. This reduces the memory footprint at
## the cost of runtime efficiency.
cacheEnabled = true
proc disableCache*() {.rtl, extern: "nro$1".} =
## the cache is discarded and disabled. The GC will reuse its used memory.
cache = nil
cacheEnabled = false
proc `&`*(a, b: Rope): Rope {.rtl, extern: "nroConcRopeRope".} =
## the concatenation operator for ropes.
if a == nil:
result = b
elif b == nil:
result = a
else:
result = newRope()
result.length = a.length + b.length
result.left = a
result.right = b
proc `&`*(a: Rope, b: string): Rope {.rtl, extern: "nroConcRopeStr".} =
## the concatenation operator for ropes.
result = a & rope(b)
proc `&`*(a: string, b: Rope): Rope {.rtl, extern: "nroConcStrRope".} =
## the concatenation operator for ropes.
result = rope(a) & b
proc `&`*(a: openArray[Rope]): Rope {.rtl, extern: "nroConcOpenArray".} =
## the concatenation operator for an openarray of ropes.
for i in countup(0, high(a)): result = result & a[i]
proc add*(a: var Rope, b: Rope) {.rtl, extern: "nro$1Rope".} =
## adds `b` to the rope `a`.
a = a & b
proc add*(a: var Rope, b: string) {.rtl, extern: "nro$1Str".} =
## adds `b` to the rope `a`.
a = a & b
proc `[]`*(r: Rope, i: int): char {.rtl, extern: "nroCharAt".} =
## returns the character at position `i` in the rope `r`. This is quite
## expensive! Worst-case: O(n). If ``i >= r.len``, ``\0`` is returned.
var x = r
var j = i
if x == nil: return
while true:
if x != nil and x.data.len > 0:
if j < x.data.len: return x.data[j]
return '\0'
else:
if x.left.length > j:
x = x.left
else:
x = x.right
dec(j, x.len)
iterator leaves*(r: Rope): string =
## iterates over any leaf string in the rope `r`.
if r != nil:
var stack = @[r]
while stack.len > 0:
var it = stack.pop
while it.left != nil:
assert(it.right != nil)
stack.add(it.right)
it = it.left
assert(it != nil)
yield it.data
iterator items*(r: Rope): char =
## iterates over any character in the rope `r`.
for s in leaves(r):
for c in items(s): yield c
proc write*(f: File, r: Rope) {.rtl, extern: "nro$1".} =
## writes a rope to a file.
for s in leaves(r): write(f, s)
proc write*(s: Stream, r: Rope) {.rtl, extern: "nroWriteStream".} =
## writes a rope to a stream.
for rs in leaves(r): write(s, rs)
proc `$`*(r: Rope): string {.rtl, extern: "nroToString".} =
## converts a rope back to a string.
result = newStringOfCap(r.len)
for s in leaves(r): add(result, s)
proc `%`*(frmt: string, args: openArray[Rope]): Rope {.
rtl, extern: "nroFormat".} =
## `%` substitution operator for ropes. Does not support the ``$identifier``
## nor ``${identifier}`` notations.
var i = 0
var length = len(frmt)
result = nil
var num = 0
while i < length:
if frmt[i] == '$':
inc(i)
case frmt[i]
of '$':
add(result, "$")
inc(i)
of '#':
inc(i)
add(result, args[num])
inc(num)
of '0'..'9':
var j = 0
while true:
j = j * 10 + ord(frmt[i]) - ord('0')
inc(i)
if frmt[i] notin {'0'..'9'}: break
add(result, args[j-1])
of '{':
inc(i)
var j = 0
while frmt[i] in {'0'..'9'}:
j = j * 10 + ord(frmt[i]) - ord('0')
inc(i)
if frmt[i] == '}': inc(i)
else: raise newException(ValueError, "invalid format string")
add(result, args[j-1])
else: raise newException(ValueError, "invalid format string")
var start = i
while i < length:
if frmt[i] != '$': inc(i)
else: break
if i - 1 >= start:
add(result, substr(frmt, start, i - 1))
proc addf*(c: var Rope, frmt: string, args: openArray[Rope]) {.
rtl, extern: "nro$1".} =
## shortcut for ``add(c, frmt % args)``.
add(c, frmt % args)
when not defined(js) and not defined(nimscript):
const
bufSize = 1024 # 1 KB is reasonable
proc equalsFile*(r: Rope, f: File): bool {.rtl, extern: "nro$1File".} =
## returns true if the contents of the file `f` equal `r`.
var
buf: array[bufSize, char]
bpos = buf.len
blen = buf.len
for s in leaves(r):
var spos = 0
let slen = s.len
while spos < slen:
if bpos == blen:
# Read more data
bpos = 0
blen = readBuffer(f, addr(buf[0]), buf.len)
if blen == 0: # no more data in file
result = false
return
let n = min(blen - bpos, slen - spos)
# TODO There's gotta be a better way of comparing here...
if not equalMem(addr(buf[bpos]),
cast[pointer](cast[int](cstring(s))+spos), n):
result = false
return
spos += n
bpos += n
result = readBuffer(f, addr(buf[0]), 1) == 0 # check that we've read all
proc equalsFile*(r: Rope, filename: string): bool {.rtl, extern: "nro$1Str".} =
## returns true if the contents of the file `f` equal `r`. If `f` does not
## exist, false is returned.
var f: File
result = open(f, filename)
if result:
result = equalsFile(r, f)
close(f)
new(N) # init dummy node for splay algorithm
{.pop.}