Last-modified: 2002 March 22
Mule bogusly considers the various ISO-8859 extended character sets as
disjoint, when ISO 8859 itself clearly considers them to be subsets of
a larger character set. For example, all of the Latin character sets
include NO-BREAK SPACE at code point 32 (ie, 0xA0 in an 8-bit code),
but the Latin-1 and Latin-2 NO-BREAK SPACE characters are considered
to be different by Mule, an obvious absurdity.
This package provides functions which determine the list of coding
systems which can encode all of the characters in the buffer, and
translate to a common coding system if possible.
***** Basic usage:
To set up the package, simply put
in your init file.
Get the latin-unity module and build as usual.
o If a buffer contains only ASCII and ISO-8859 Latin characters, the
buffer can be "unified", that is treated so that all characters are
translated to one charset that includes them all. If the current
buffer coding system is not sufficient, the package will suggest
alternatives. It prefers ISO-8859 encodings, but also suggests
UTF-8 (if available; 21.4+ feature, currently requires Mule-UCS),
ISO 2022 7-bit, or X Compound Text if no ISO 8859 coding system is
It allows the user to use other coding systems, and the list of
suggested coding systems is Customizable.
This probably also is useful out of the box if the buffer contains
non-Latin characters in addition to a mixture of Latin characters.
For example, it would reduce a buffer originally encoded in
ISO-2022-JP (including Latin-1 characters) to ISO 8859/1 if all
the Japanese were deleted. (untested)
o ISO 8859/15 for XEmacs 21.4 (lightly tested) and 21.1 (untested),
including binding the EuroSign keysym to ISO 8859/15 0xA4.
To get 'iso-8859-15 preferred to 'iso-8859-1 in autodetection, use
(set-coding-category-system 'iso-8-1 'iso-8859-15). (untested)
Alternatively set language environment to Latin-9.
If all you want is ISO 8859/15 support, you can either copy the
ISO 8859/15 setup to another file, or `(require 'latin-unity-vars)'
and `(require 'latin-euro-input)'.
o Hooks into `write-region' to prevent (or at least drastically
reduce the probability of) introduction of ISO 2022 escape
sequences for "foreign" character sets. This hook is not set by
default in this package yet; try M-x latin-unity-example RET for a
short introduction and some useful C-x C-e'able exprs.
This may permit us to turn off support for those sequences
entirely in our ISO 8859 coding-systems.
o Interactive functions to _remap_ a region between character sets
(preserving character identity) and _recode_ a region (preserving
the code point). The former is probably not useful if the
automatic function is working at all, but provided for
completeness. The latter is useful if Mule mistakenly reads an
ISO 8859/2 file as ISO 8859/1; you can change it without rereading
the file. Since it's region-oriented, you can also deal with cut
and paste from dumb applications that export everything as ISO 8859/1.
o A nearly comprehensive Texinfo manual contains a discussion of
why these things happen, why they can't be 100% avoided in an 8-bit
world, and some defensive measures users can take, as well as the
usual install, configure, and operating instructions.
o latin-unity itself depends only on mule-base in operation. Table
generation depends on Unicode support such as Mule-UCS or Ben's
ben-mule-21-5 workspace, and the package build currently requires
Mule-UCS. The input method depends on LEIM and fsf-compat.
o Doesn't automatically save pure ASCII files in ASCII superset
encodings like iso-2022-jp. Workaround: put an ISO 8859 coding
system in `latin-unity-preapproved-coding-system-list'.
o Need `(require 'latin-euro-input)' to get Quail support.
o Possible performance hit on large (> 20kB) buffers with many
(>20%) non-ASCII characters. Partially optimized, but see near
latin-unity.el for possible further optimizations.
o Package depends on Mule-UCS, LEIM (Quail), and fsf-compat.
o This README is too long.
o Must load latin-unity-vars before reading a file with ISO 8859/15,
there is no way to autoload a charset. (Can't be fixed without
changing XEmacs core.)
Planned, mostly near future:
o Fix the misfeatures.
o Check -*- coding: codesys -*- cookies for consistency.
o Fix JIS Roman (as an alternative to ASCII) support.
o Support Latin-10 (ISO 8859/16) and maybe others.
o More UI features (like highlighting unrepresentable chars in buffer).
o Integration to development tree (but probably not 21.4, this
package should be good enough).
o Hook into MUAs.
o GNU Emacs support.
Not planned any time soon:
o Extend to process buffers in some way, which looks very hard.
o Han-unity. This is not entirely analogous to Latin unity, and
needs to be treated very carefully.
latin-unity.el is the main library, providing the detection and translation
functionality, including a hook function to hang on `write-region-pre-hook'.
latin-unity-vars.el contains the definition of variables common to
latin-unity-latin9.el defines ISO 8859/15 and the Latin-9 environment.
latin-euro-input.el contains Dave Love's Quail input method for Latin 9.
latin-unity-tables.el contains the table of feasible character sets and
equivalent Mule characters from other character sets for the various Mule
representations of each character. Automatically generated. Used only when
Unicode support is not present.
latin-unity-utils.el contains utilities for creating the equivalence table,
and dumping it to a file. Used in preference to the precomputed table when
Unicode support is available.