+Thu Apr 23 17:04:59 1992 Per Hammarlund (perham at perham.nada.kth.se)
+ * Stepped up the version number, 0.9.6.
+ * Added edict-add-word, which basically initiates a private
+ dictionary session without starting with a word.
+Wed Apr 22 09:30:14 1992 Per Hammarlund (perham at perham.nada.kth.se)
+ * Changed the declaration of the *edict-rmoaji-remaps* to be a
+ defvar followed by a setq.
+ * Rewrote and corrected some of the short getting started guide at
+ the beginning of edict.el.
+Mon Apr 6 06:58:59 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Add eval-when around defstruct. The compiler half of
+ eval-when is in a patch that is loading into the compiler
+ * Make the english-plural rule exclude ies.
+ * Add edict-plural-ies.
+ * Call edict-expand-string with the right arguments.
+Fri Apr 3 20:38:03 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Add english-plurals rule.
+Thu Apr 2 02:18:27 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Add rules for �����Ǥ� and their conjucations.
+ * forgot to exclude newlines in edict-slash-internal.
+ * forgot to exclude newlines in edict-char-bracket.
+ * Forgot to exclude newlines in edict-line-has-english.
+ * Forgot to exclude newlines in edict-parse-entry.
+ * Forgot to exclude newlines in pattern for finding english defs
+ in edict-insert-english.
+Wed Apr 1 13:28:18 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Put the cursor at the right place when creating an entry and
+ * A bunch of minor tuneups to various things...
+ * Make the dictionary-editing commands go through a common wrapper
+ to keep the electric stuff consistent.
+ * Add similar command for /.
+ * Add edict-open-bracket, edict-close-bracket, in dictionary
+ * Allow edict-new-entry to take an argument. <= 4 copies the
+ kanji word into the new entry; >4 (i.e. c-U c-U) copies both
+ the word and definition, allowing you to enter an alternate
+ * Split out edict-insert-entry from edict-add-entry-to-file,
+ and add an inverse: edict-parse-entry.
+ * Allow a list of translations in edict-add-entry-to-file.
+ * Allow windows to be as small as emacs will allow. This is
+ important, because when we look for a suitably-sized window,
+ this is all the bigger we look for.
+ * Add some more comments, and convert and/or/one-armed-if
+ expressions into when/unless, since those are now provided
+ by the (require 'cl). This makes a number of things more
+ * Move edict-test-rule and edict-test-string to edict-test.el.
+ * Merge in Per's changes below.
+ * Add a bit more commentary.
+ * Fix mising "w" and "W" in remap table.
+Tue Mar 31 06:16:26 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * edict-subst-affix's args can now be any function of one argument
+ (the matched substring).
+ * Optimize edict-ignore in edict-subst-modified-affix, and
+ add it to edict-subst-affix.
+ * Add ��ʤ� rule to try looking up number expressions without
+Mon Mar 30 10:16:21 1992 Per Hammarlund (perham at perham.nada.kth.se)
+ * Made sure that edict-new-entry put a closing '/' at the end of
+ the current row. This only works when the user does
+ edict-new-entry, not for save. Sigh.
+ * Up version number 0.9.5
+ * Added handling of SKK in edict-set-henkan. Crude
+Mon Mar 30 00:19:05 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Don't insert  after hiragana-only words, either!
+ * Electric dictionary editing: When adding entries,
+ Tab now moves between fields, and Return starts a new entry
+ on the next line. It arranges to be in nihongo mode when
+ appropriate. Also, it no longer inserts  after katakana
+Sun Mar 29 14:03:09 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Replace (eval (cons 'and kludge with every.
+ * filter out some ��->�Ƥ� in ���ʤ�conjunctive
+ * Optimize edict-identity in rule RHS's
+ * Use regexp character categories where possible.
+ * fix typo: window-next -> next-window.
+Fri Mar 27 02:02:58 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Get around nemacs regexp bug that was still deleting the katakan
+ long-sound dash even after I fixed the regexp. (Off-by-one
+ error on the handling of the end of a range. Sigh.)
+Thu Mar 26 20:56:00 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Whitespace list accidentally included the katakana long-sound dash.
+ * The conjunctive rules patterns only allowed kanji, not
+Wed Mar 25 17:57:09 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Merge in Per's change to edict-eigo-one-word.
+ * Up the version to 0.9.4
+ * Make whitespace stuff include all the different funny chars.
+ * Show the dictionary version number(s) after loading them.
+ * Replace '(lambda ...) with (function (lambda ...)).
+ * Make the default edict pathname be just "edictj", so it searches
+ load path to find it. This should minimize installation
+ headaches. You can still override this in your .emacs file.
+ * Use load-path to search for edict files if they're relative.
+ * Handle �� and punctuation in kanji insertion. Also, let a
+ couple of rules work with katakana stems (i.e. ����-derived stuff).
+Mon Mar 23 11:18:24 1992 Per Hammarlund (perham at perham.nada.kth.se)
+ * Fixed a bug that caused it not to find a english word at the
+ beginning of the buffer. It ran backward over (point-min).
+ Change made in "edict-eigo-one-word", >= to >.
+Mon Mar 23 07:37:03 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Increment version to 0.9.3
+ * Add edict-insert, edict-insert-english, edict-insert-���ܸ�.
+ The main one to use is edict-insert, which figures out what
+ language to insert based on the last lookup.
+ * Disallow the �� -> �� -> �� path.
+ * For the various �ޤ� rules, filter out cases derived from �ޤ���.
+ * Don't apply the potential-form rules if there's just the suffix.
+ * ���ƻ졼�� -- Be a whole lot more careful about what we consider
+ an adverb/adjective, based on the previous form. This should
+ eliminate a lot of ������ verbs from having spurious forms.
+ * I decided that this filtering scheme isn't so bad after all,
+ if the decision is made on just the preceding version.
+ Having parts of speech would still be better, since it would
+ localize the knowledge of what part of speech an ending could
+ be. But I think this will do for now.
+ * Separate out (��|��|���)�� case from ��ư�졼�� to ��ư�졼����,
+ because this case must not include ��. ���� should not become
+ ��. ���� is just a non-grammatical conversational case, anyway,
+ and only occurs after �� so far as I know.
+ * Add more filtering to problematic conjunctive forms.
+ * Add filtering to filter out rules which are locally plausible,
+ but could not be applied if what was removed is taken into
+ consideration. eg �Ԥ� -> �Ԥ� -> �Ԥ� is not legit, even
+ though each individual step is, because the first step is based
+ on the assumption that it's an adverb, and turns it into an
+ adjective. The second assumes its the conjunctive form of a
+ verb, and tries to find the dictionary form.
+ This first pass is pretty bogus. What it does is try to deduce
+ it from what endings it has had along the way. This can't
+ really work, because there's no reasonable way to determine
+ whether the ending in question is applying to this root, or
+ or some further extention of this root.
+ The right thing, is for rules to filter by/assert a part of speach.
+ But first, I'm going to try this, just to get things going.
+ * Remove redundant rule: �֤����פ����̥롼��
+ * Merge in Per's changes. Up the version to 0.9.2
+ * Provide error checking for unrooted patterns, so that last
+ bug can't happen again.
+ * Fix �ֲ����פ�����conjunctive rules to only match at the end.
+Sat Mar 21 10:53:51 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Eliminate duplicates in the result when we have substring matches.
+ * Potential form's rules were completely mangled; they were all
+ clones of the same version, as if incomplete edited or
+ something. I fixed them up, and added another test to the suit.
+ The test suite happened to test the one case that the rest were
+ cloned from, and ����ư��, which worked.
+ * Anti-infinite-recursion check didn't catch quite all the
+ possible ways. Also, in similar circumstances, it could
+ put duplicates into the result.
+ * Add �� to other-suffixes, and Ⱦ to other-prefixes.
+Fri Mar 20 11:14:00 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Remove punctuation before lookup.
+ * Make the windowing code more robust when editing in a tiny
+Thu Mar 19 15:03:52 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Canonicalize number expressions.
+Tue Mar 17 10:35:45 1992 Per Hammarlund (perham at perham.nada.kth.se)
+ * Added to a regexp in edict-display so that english verbs that
+ have a dictionary entry like /to XXX/ shall be treated as exact
+ matches in the match-list.
+ * fixed a small index error in edict-remap-romaji.
+Tue Mar 17 10:06:48 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Merged in my changes below into 0.9, and made it 0.9.1
+Mon Mar 16 00:28:23 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Prevent infinite recursion if rules have a loop.
+ * Add conjunctive form. (�ꤤ�����ꤤ���Ӥ�), and ���롿���롿��
+Mon Mar 16 09:38:44 1992 Per Hammarlund (perham at perham.nada.kth.se)
+ * Set the version to 0.9 and the date to current.
+ * Added functions to clean up a eigo string that, for whatever
+ reason has been written in both ASCII and JIS/EUC. For instance
+ the string "str�����", will be tranformed to plain "strange".
+ The remap relation can be done better, and it should probably be
+ more complete. How about a complete new nemacs-string* function
+ * Changed the function edict-find-word-at-point that finds an
+ english word to take eigo written in Japanese [��-����-��] into
+ * Rewrote the version code, more things to be able to keep the
+ date and the version on different variables.
+ * Added/rewrote the "Getting Started Guide" in the file edict.el.
+ * Started on the documentation. edict.texinfo, Took out a number
+ of lines from the edict.el file.
+ * Changed the name of the variable *private-edict-file* to
+ *edict-private-file*, just to keep variable names consistent.
+Mon Mar 16 00:28:23 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Add �����ޡ��͡�ĺ���������������
+ * Add ���� to ��ư�졼����
+ * We have to add 1 to the window height because they include the
+ mode line. This was hiding the last entry found when there were
+ * ��ư�� ���� ��ư�졼���Ľ�ư�졼���ޤ�ʬ���Ƥ��������ֺ����夲���
+ * Add the ���� verb endings as alternates to ���ʤ�, and
+Sun Mar 15 00:01:05 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * �֤��ơ����ơפ����̥롼��: Fix bug introduced in fixing another
+ * Add �� to the list of endings that ��ư�� apply to. This is
+ actually confusing things a bit, but in a way that helps the user.
+ Logically, if an adverb appears before a verb which is sometimes a
+ ��ư��, it probably is modifying the verb, not the other way
+ around. But the user was probably asking about the adverb, since
+ the ��ư�� are all very basic words the user probably knows. I
+ suppose I could start returning ��ư�� in addition to the main
+ verb, so the user would see these as well. I suspect this would
+ be more distraction than benefit. If the user doesn't know the
+ part that was stripped off, he can just mark it, and inquire.
+ * Add �ʤ� as ��ư��. Add �������Ρ����פȡֹԤ��פȡ־夲���
+ * Be more careful about ordering the results.
+ * Remove debugging notify accidentally left inserted.
+ * Allow modifiers rule after ����ư��.
+ * Correct an error in �֤���פ����̥롼��.
+ * Suppress single-kanji matches unless they're needed.
+ * Didn't have anything for handling ������!!
+ * Forgot to drop the ���� from humble form.
+ * Fix up adjectives once more.
+ * �ɸ����Ƭ�� tried to be both prefix and suffix(!)
+ * �Ƥ��ޤ� and �Ǥ��ޤ� cant be combined after all.
+ * Fix incorrect kanji in other-prefixes.
+ * Coallesce the �Ƥ��ޤ� family into a single rule, and make it
+ properly preserve the �ơ���� distinction. No need for
+ past-tense forms since they're regular; the past-tense rules
+ * Apply the prefixes only to ����
+ * Added ���� to suffixes.
+ * Add basic affixes used with ̾�졣
+ * Explicit rules for ���ʤ����ʤ���
+ * Get ���ƻ� negative right, and include more formal varients.
+ * ʿ��̾�Ρ֤��� as honorific prefix, in addition to ��.
+ * Remove ���� from ��ư��; that gets covered elsewise anyway.
+ * Minimum length on ��ư��.
+ * Only do command if at least two chars long.
+ * Tighten up a couple more rules wrt ���� only after ���� or ����.
+ * Tighten up various rules wrt legal verb endings. No �š���, etc.
+ * Typo in �֤��褦�פ����̥롼��, ����褦�ס��֤��褦��
+ * Forgot a few '$'s on the ends of patterns.
+ * Add a complete set of volitionals. This pretty much completes
+ * Supply a few more ��ư�� and other endings.
+ * Irregular �褿���롢�������롢��������
+ * Avoid generating �� or �� as entire words.
+ * Fix up ��������Ρ֤����פ������� -> ���ʤΡ֤����פ�������
+ * Add an ����ư�� version of the desiderative rule.
+ * Added �������� to desideratives.
+ * Tightened up adjective rules so they don't go off on ���롣
+ * Added Irregular ��ޤ������ޤ���
+ * Add irregular info for ���ʤ�, �ʤ�.
+ * Correct mistaken rule for ���ʤ�.
+Sat Mar 14 16:22:07 1992 Bob Kerns (rwk at taunton.crl.dec.com)
+ * Add edict-test-string.
+ * Add debugging feature to edict-expand-string to help track down
+ what rules do what in the event of surprises.
+ * Handle ���� as part of the �ơ��� rules. Faster, and doesn't
+ generate spurious instance of plain past.
+ * Include the zenkaku versions of the punctuation in
+ * Clean up some lurking problems in edict-clean-up-(kanji|eigo)
+ * Fix a problem with exact matches in the substitution routines.
+ * Make edict-expand-string more aggressive in preventing possible
+ * Add ����� rule for its irregularity.
+ * Add ����� rule for its irregularity.
+ * Add �Ԥ� special rule for its irregularity
+ * Additions to improve window handling
+ Add private dictionary, and commands to add to it.
+ Add morphologic analysis so that the dictionary forms
+ of words can be found from inflections/affixes/etc.
+ Add rule for plain negative, plain imperitive.
+ Add rule for desiderative.
+ Make it require cl.el, since I got tired of using a crippled lisp.
+ Use DEFSTRUCT for the rules, and give them names so that
+ you can redefine a rule without having to reload the entire
+ Add '#' to the characters that get flushed from kanji words
+ since that's a common inclusion character in fj newsgroups.
+ Split the set out into a variable *edict-kanji-whitespace*
+ to aid in user extensibility.
+ Added rule for ����ư�� �ơ��� -> �� conversion, which
+ I accidentally deleted earlier.
+ Added conditional form for adjectives.
+ Added causitive, passive, ����ư�� potential
+ Added some more missing ����ư�� rules, and
+ tightened up a number of rules, especially ����ư�� rules.
+ Upgrade the sorting of exact matches to the top to
+ handle exact matches of dictionary forms produced by
+ the rule set. For example, ���ʤ��� should have ����
+ sorted to the top of the list.
+ Updated credits. Added *edict-eigo-whitespace* like
+ *edict-kanji-whitespace*, and fixed a bug in the code.
+ Added rules for �����㤦��������ä�
+Fri Mar 13 17:45:27 1992 Per Hammarlund (perham at perham.nada.kth.se)
+ * Merged Bob's extensions just below into the distribution
+ version. (Some of the things mentioned right above here,
+Fri Mar 13 17:44:06 1992 Bob Kerns (rwk at crl.dec.com)
+ * Added functionality to translate verbs and adjectives to a basic
+Wed Mar 11 09:41:23 1992 Per Hammarlund (perham at perham.nada.kth.se)
+ * Added edict-version and a variable *edict-version* to keep track
+ of the current version of the software.
+ * Type checks on the variable *edict-files* in function
+Tue Mar 10 10:22:21 1992 Bob Kerns (rwk at crl.dec.com)
+ * I've fixed some lingering bugs in restoring the window sizes
+ * I've made it canonicalize whitespace, so that you can select
+ jukugo that wrap around lines, include whitespace, or whatever.
+ Also, multi-word english phrases can include newlines,
+Tue Mar 10 07:11:24 1992 Bob Kerns (rwk at crl.dec.com)
+ * I fixed the displaying to not screw up if there's a window above
+ (for example, when using it on fj.* gnus articles it used to
+ gradually grow the Subject window and shrink the Article window).
+ * I put giving a positive argument to the searching commands add a
+ new entry. Locally, I put these commands on m-& and m-*; and if
+ m-& doesn't find a compound, I do c-U m-& and add it.
+ * m-- m-& (or whatever) will remove the window. I did this
+ because c-X 1 is too much; if I was in 2 window mode, I don't want
+ to go to 1 window mode.
+ * I made the display show any exact matches before the inexact
+ matches. So if I do it on ��, I get that line first, instead of
+ buried in amongst all the compounds that sort earlier.
+Mon Mar 9 00:00:00 1992 Bob Kerns (rwk at crl.dec.com)
+ * This gives you m-X edict-add-english and m-X
+ edict-add-kanji, and automagically maintains a private
+ dictionary in *edict-private-file*.
+1991, 1992 Per Hammarlund (perham at perham.nada.kth.se)
+ * Wrote the bare bones of edict.el.