Add CJK codecs support as discussed on python-dev. (SF #873597)
Several style fixes are suggested by Martin v. Loewis and
Marc-Andre Lemburg. Thanks!
diff --git a/Modules/cjkcodecs/README b/Modules/cjkcodecs/README
new file mode 100644
index 0000000..dc32574
--- /dev/null
+++ b/Modules/cjkcodecs/README
@@ -0,0 +1,87 @@
+Notes on cjkcodecs
+-------------------
+This directory contains source files for cjkcodecs extension modules.
+They are based on CJKCodecs (http://cjkpython.i18n.org/#CJKCodecs)
+as of Jan 17 2004 currently.
+
+
+
+To generate or modify mapping headers
+-------------------------------------
+Mapping headers are imported from CJKCodecs as pre-generated form.
+If you need to tweak or add something on it, please look at tools/
+subdirectory of CJKCodecs' distribution.
+
+
+
+Notes on implmentation characteristics of each codecs
+-----------------------------------------------------
+
+1) Big5 codec
+
+ The big5 codec maps the following characters as cp950 does rather
+ than conforming Unicode.org's that maps to 0xFFFD.
+
+ BIG5 Unicode Description
+
+ 0xA15A 0x2574 SPACING UNDERSCORE
+ 0xA1C3 0xFFE3 SPACING HEAVY OVERSCORE
+ 0xA1C5 0x02CD SPACING HEAVY UNDERSCORE
+ 0xA1FE 0xFF0F LT DIAG UP RIGHT TO LOW LEFT
+ 0xA240 0xFF3C LT DIAG UP LEFT TO LOW RIGHT
+ 0xA2CC 0x5341 HANGZHOU NUMERAL TEN
+ 0xA2CE 0x5345 HANGZHOU NUMERAL THIRTY
+
+ Because unicode 0x5341, 0x5345, 0xFF0F, 0xFF3C is mapped to another
+ big5 codes already, a roundtrip compatibility is not guaranteed for
+ them.
+
+
+2) cp932 codec
+
+ To conform to Windows's real mapping, cp932 codec maps the following
+ codepoints in addition of the official cp932 mapping.
+
+ CP932 Unicode Description
+
+ 0x80 0x80 UNDEFINED
+ 0xA0 0xF8F0 UNDEFINED
+ 0xFD 0xF8F1 UNDEFINED
+ 0xFE 0xF8F2 UNDEFINED
+ 0xFF 0xF8F3 UNDEFINED
+
+
+3) euc-jisx0213 codec
+
+ The euc-jisx0213 codec maps JIS X 0213 Plane 1 code 0x2140 into
+ unicode U+FF3C instead of U+005C as on unicode.org's mapping.
+ Because euc-jisx0213 has REVERSE SOLIDUS on 0x5c already and A140
+ is shown as a full width character, mapping to U+FF3C can make
+ more sense.
+
+ The euc-jisx0213 codec is enabled to decode JIS X 0212 codes on
+ codeset 2. Because JIS X 0212 and JIS X 0213 Plane 2 don't have
+ overlapped by each other, it doesn't bother standard conformations
+ (and JIS X 0213 Plane 2 is intended to use so.) On encoding
+ sessions, the codec will try to encode kanji characters in this
+ order:
+
+ JIS X 0213 Plane 1 -> JIS X 0213 Plane 2 -> JIS X 0212
+
+
+4) euc-jp codec
+
+ The euc-jp codec is a compatibility instance on these points:
+ - U+FF3C FULLWIDTH REVERSE SOLIDUS is mapped to EUC-JP A1C0 (vice versa)
+ - U+00A5 YEN SIGN is mapped to EUC-JP 0x5c. (one way)
+ - U+203E OVERLINE is mapped to EUC-JP 0x7e. (one way)
+
+
+5) shift-jis codec
+
+ The shift-jis codec is mapping 0x20-0x7e area to U+20-U+7E directly
+ instead of using JIS X 0201 for compatibility. The differences are:
+ - U+005C REVERSE SOLIDUS is mapped to SHIFT-JIS 0x5c.
+ - U+007E TILDE is mapped to SHIFT-JIS 0x7e.
+ - U+FF3C FULL-WIDTH REVERSE SOLIDUS is mapped to SHIFT-JIS 815f.
+