Package: fonttools
Version: 1.99+2.0b1+cvs20031014-1
Severity: normal
Tags: patch
ttx accidentally uses a O(n^2) algorithm for converting hexdump to binary:
hexdata = ""
for chunk in content:
if type(chunk) == type(""):
hexdata = hexdata + chunk
This *copies* hexdata on every line. The mistake is understandable,
as the code looks similar to perl's (and few other languages') idiom
for $chunk(@content) { $hexdata .= $chunk if something($chunk) }
that works in place.
For Western fonts, that are usually in ~100kB size range, the O(n^2)
behaviour is not very problematic, but the CJK fonts are often 5-10MB
or larger, so it takes about forever for the hexdump to binary conversion
on ttx->ttf to finish (I don't know how long exactly, I ^C-ed the program
and started debugging after about 25 minutes on 6MB Sazanami Gothic font)
The patch replaces 2 instances of:
data = ""
for:
data = data + something
return data
with semantically equivalent, but O(n):
data = []
for:
data.append(something)
return "".join(data)
Actually I converted readHex to more pythonesque list-comprehension format
instead of appending in loop (it's equivalent to the loop, after compilation),
but this isn't very important.
In a different thread,
fonttools seems to have many other bugs that appear when I try to process CJK
fonts
with it, and the last release on sf.net was in 2002. Is it abandoned ?
diff -ru fonttools-1.99+2.0b1+cvs20031014.orig/Lib/fontTools/misc/textTools.py
fonttools-1.99+2.0b1+cvs20031014/Lib/fontTools/misc/textTools.py
--- fonttools-1.99+2.0b1+cvs20031014.orig/Lib/fontTools/misc/textTools.py
2000-02-01 16:28:18.000000000 +0100
+++ fonttools-1.99+2.0b1+cvs20031014/Lib/fontTools/misc/textTools.py
2005-06-06 16:45:46.000000000 +0200
@@ -11,11 +11,7 @@
def readHex(content):
"""Convert a list of hex strings to binary data."""
- hexdata = ""
- for chunk in content:
- if type(chunk) == type(""):
- hexdata = hexdata + chunk
- return deHexStr(hexdata)
+ return deHexStr(''.join([ chunk for chunk in content if
isinstance(chunk,str) ]))
def deHexStr(hexdata):
"""Convert a hex string to binary data."""
@@ -23,10 +19,11 @@
hexdata = string.join(parts, "")
if len(hexdata) % 2:
hexdata = hexdata + "0"
- data = ""
+ data = []
for i in range(0, len(hexdata), 2):
- data = data + chr(string.atoi(hexdata[i:i+2], 16))
- return data
+ data.append(chr(string.atoi(hexdata[i:i+2], 16)))
+ return "".join(data)
+
def hexStr(data):
"""Convert binary data to a hex string."""
