latin-prefix input method niggles

latin-prefix input method niggles

[Robert Pluim]

The description of the latin-prefix input method contains

       cedilla   |   ~    | ~c -> ç
        breve    |   ~    | ~a -> ă
        caron    |   ~    | ~c -> č
      dot above  | ~ / .  | ~o -> ġ   /o -> ġ   .o -> ġ

Problems:

1. the 'breve' description is inaccurate, actually ~a -> ã [1]
2. the 'caron' description is inaccurate, actually ~c -> ç

There is however no definition for creating 'a with breve' nor 'c with
caron' in that input method

3. the dot above entry is also inaccurate, actually ~o -> õ  /o -> ø .o -> .o

There are definitions for /g -> ġ and .g -> ġ , however

Proposed solution:

1. I donʼt have one. Other input methods use ~a for either a with
tilde or a with breve. We could just fix the docstring

2. We could do ~c -> č instead of ç , which is quite common in other
input methods, and add ,c -> ç , which is in line
with latin-postfix

3. This is a doc issue, we could replace that line with:

  dot above  |  / .   | /g -> ġ   .g -> ġ

Thoughts? Even "donʼt touch this, ever, people are wedded to these
nits" :-)

Robert

Footnotes:
[1]  In fact I just noticed a similar doc error in the
     romanian-alt-prefix method

Re: latin-prefix input method niggles

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Date: Thu, 10 Jan 2019 19:39:29 +0100
> 
> The description of the latin-prefix input method contains
> 
>        cedilla   |   ~    | ~c -> ç
>         breve    |   ~    | ~a -> ă
>         caron    |   ~    | ~c -> č
>       dot above  | ~ / .  | ~o -> ġ   /o -> ġ   .o -> ġ
> 
> Problems:
> 
> 1. the 'breve' description is inaccurate, actually ~a -> ã [1]
> 2. the 'caron' description is inaccurate, actually ~c -> ç
> 
> There is however no definition for creating 'a with breve' nor 'c with
> caron' in that input method
> 
> 3. the dot above entry is also inaccurate, actually ~o -> õ  /o -> ø .o -> .o
> 
> There are definitions for /g -> ġ and .g -> ġ , however

IMO, these are bugs, not just documentation issues.

> Proposed solution:
> 
> 1. I donʼt have one. Other input methods use ~a for either a with
> tilde or a with breve. We could just fix the docstring
> 
> 2. We could do ~c -> č instead of ç , which is quite common in other
> input methods, and add ,c -> ç , which is in line
> with latin-postfix
> 
> 3. This is a doc issue, we could replace that line with:
> 
>   dot above  |  / .   | /g -> ġ   .g -> ġ
> 
> Thoughts?

I think we should make latin-prefix be consistent with latin-postfix.
Each one of the above is correct in the latter, AFAICT, so we should
just fix the problems in latin-prefix.  E.g., ~a should produce both ã
and ă, ~c should produce č, and ġ should be produced by .g.  I see no
reason to have these two behave inconsistently; do you?

Re: latin-prefix input method niggles

[Robert Pluim]

Eli Zaretskii <eliz@gnu.org> writes:

> I think we should make latin-prefix be consistent with latin-postfix.
> Each one of the above is correct in the latter, AFAICT, so we should
> just fix the problems in latin-prefix.  E.g., ~a should produce both ã
> and ă, ~c should produce č, and ġ should be produced by .g.  I see no
> reason to have these two behave inconsistently; do you?

.g already produces ġ, changing ~c to produce č is easy. I take it
youʼre ok with adding ,c -> ç

*time passes* I hadn't realised you could define multiple productions
 for the same mapping. So we instead end up with

~c -> ç
~c -> č
~a -> ã
~a -> ă

in that order, since that keeps the current latin-prefix mappings of
~c and ~a first. We could also add ,c -> ç for consistency with
latin-postfix.

Robert 

Re: latin-prefix input method niggles

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Cc: emacs-devel@gnu.org
> Date: Thu, 10 Jan 2019 21:36:01 +0100
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > I think we should make latin-prefix be consistent with latin-postfix.
> > Each one of the above is correct in the latter, AFAICT, so we should
> > just fix the problems in latin-prefix.  E.g., ~a should produce both ã
> > and ă, ~c should produce č, and ġ should be produced by .g.  I see no
> > reason to have these two behave inconsistently; do you?
> 
> .g already produces ġ, changing ~c to produce č is easy. I take it
> youʼre ok with adding ,c -> ç

Yes, because it'd be consistent with latin-postfix.

> *time passes* I hadn't realised you could define multiple productions
>  for the same mapping. So we instead end up with
> 
> ~c -> ç
> ~c -> č
> ~a -> ã
> ~a -> ă
> 
> in that order, since that keeps the current latin-prefix mappings of
> ~c and ~a first. We could also add ,c -> ç for consistency with
> latin-postfix.

Right.

Re: latin-prefix input method niggles

[Robert Pluim]

Robert Pluim <rpluim@gmail.com> writes:

>>> *time passes* I hadn't realised you could define multiple productions
>>>  for the same mapping. So we instead end up with
>>> 
>>> ~c -> ç
>>> ~c -> č
>>> ~a -> ã
>>> ~a -> ă
>>> 
>>> in that order, since that keeps the current latin-prefix mappings of
>>> ~c and ~a first. We could also add ,c -> ç for consistency with
>>> latin-postfix.
>>
>> Right.

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> From: Robert Pluim <rpluim@gmail.com>
>>> Cc: emacs-devel@gnu.org
>>> Date: Thu, 10 Jan 2019 21:36:01 +0100
>>> 
>>> Eli Zaretskii <eliz@gnu.org> writes:
>>> 
>>> > I think we should make latin-prefix be consistent with latin-postfix.
>>> > Each one of the above is correct in the latter, AFAICT, so we should
>>> > just fix the problems in latin-prefix.  E.g., ~a should produce both ã
>>> > and ă, ~c should produce č, and ġ should be produced by .g.  I see no
>>> > reason to have these two behave inconsistently; do you?
>>> 
>>> .g already produces ġ, changing ~c to produce č is easy. I take it
>>> youʼre ok with adding ,c -> ç
>>
>> Yes, because it'd be consistent with latin-postfix.
>>

This is what I have so far. The first two hunks correct the
documentation of romanian-prefix and romanian-alt-prefix to match the
actual implementation, which I can commit separately. The last three
hunks implement what we discussed above.

I have no idea why ~o produces ġ in latin-prefix, if thatʼs an error I
can remove it. It might be confusion caused by the docstring of
latin-3-prefix, which says

  dot above  |   / .  | /g -> ġ   .o -> ġ

when itʼs actually .g -> ġ , with no mapping for 'o' at all.

Iʼm going to stop looking now before I get stuck further down the
rabbit hole.

diff --git a/lisp/leim/quail/latin-pre.el b/lisp/leim/quail/latin-pre.el
index 63565d4ced..3b24e9ee36 100644
--- a/lisp/leim/quail/latin-pre.el
+++ b/lisp/leim/quail/latin-pre.el
@@ -320,7 +320,7 @@
 
     effect   | prefix | examples
  ------------+--------+------------------
-    tilde    |   ~    | ~a -> ă
+    breve    |   ~    | ~a -> ă
   circumflex |   ^    | ^a -> â, ^i -> î
    cedilla   |   ,    | ,s -> ş, ,t -> ţ
    ~         |   ~    | ~~ -> ~
@@ -342,11 +342,11 @@
 
     effect   | prefix | examples
  ------------+--------+------------------
-    tilde    |   \"    | \"a -> â
-  circumflex |   \\='    | \\='a -> â, \\='i -> î
-   cedilla   |   \\='    | \\='s -> ş, \\='t -> ţ
-   \\='         |   \\='    | \\='\\=' -> \\='
-   \"         |   \"    | \"\" -> \"
+    breve    |   \\='    | \\='a -> ă
+ circumflex  |  \" \\='   | \"a -> â  \\='i -> î
+   cedilla   |   \\='    | \\='s -> ş  \\='t -> ţ
+      \\='      |   \\='    | \\='\\=' -> \\='
+      \"      |   \"    | \"\" -> \"
 " nil t nil nil nil nil nil nil nil nil t)
 
 (quail-define-rules
@@ -1089,15 +1089,15 @@
 
     effect   | prefix | examples
  ------------+--------+----------
-    acute    |   \\='    | \\='a -> á, \\='\\=' -> ´
+    acute    |   \\='    | \\='a -> á  \\='\\=' -> ´
     grave    |   \\=`    | \\=`a -> à
   circumflex |   ^    | ^a -> â
   diaeresis  |   \"    | \"a -> ä  \"\" -> ¨
     tilde    |   ~    | ~a -> ã
-   cedilla   |   ~    | ~c -> ç
+   cedilla   |  , ~   | ,c -> ç  ~c -> ç
+    caron    |   ~    | ~c -> č  ~g -> ğ
     breve    |   ~    | ~a -> ă
-    caron    |   ~    | ~c -> č
-  dot above  | ~ / .  | ~o -> ġ   /o -> ġ   .o -> ġ
+  dot above  | ~ / .  | ~o -> ġ   /g -> ġ   .g -> ġ
     misc     | \" ~ /  | \"s -> ß  ~d -> ð  ~t -> þ  /a -> å  /e -> æ  /o -> ø
    symbol    |   ~    | ~> -> »  ~< -> «  ~! -> ¡  ~? -> ¿  ~~ -> ¸
    symbol    |  _ /   | _o -> º  _a -> ª  // -> °  /\\ -> ×  _y -> ¥
@@ -1252,7 +1252,10 @@
  ("~>" ?\»)
  ("~?" ?¿)
  ("~A" ?Ã)
+ ("~A" ?Ă)
  ("~C" ?Ç)
+ ("~C" ?Č)
+ (",C" ?Ç)
  ("~D" ?Ð)
  ("~G" ?Ğ)
  ("~N" ?Ñ)
@@ -1265,7 +1268,10 @@
  ("~Z" ?Ž)
  ("~`" ?˘)
  ("~a" ?ã)
+ ("~a" ?ă)
  ("~c" ?ç)
+ ("~c" ?č)
+ (",c" ?ç)
  ("~d" ?ð)
  ("~e" ?€)
  ("~g" ?ğ)

Re: latin-prefix input method niggles

[Eli Zaretskii]

> From: Robert Pluim <rpluim@gmail.com>
> Date: Fri, 11 Jan 2019 11:58:24 +0100
> 
> This is what I have so far.

LGTM, thanks.

> I have no idea why ~o produces ġ in latin-prefix, if thatʼs an error I
> can remove it. It might be confusion caused by the docstring of
> latin-3-prefix, which says
> 
>   dot above  |   / .  | /g -> ġ   .o -> ġ
> 
> when itʼs actually .g -> ġ , with no mapping for 'o' at all.

I think ~o is indeed a mistake, and should be deleted.  It makes no
sense.

Re: latin-prefix input method niggles

[Robert Pluim]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Robert Pluim <rpluim@gmail.com>
>> Date: Fri, 11 Jan 2019 11:58:24 +0100
>> 
>> This is what I have so far.
>
> LGTM, thanks.
>
>> I have no idea why ~o produces ġ in latin-prefix, if thatʼs an error I
>> can remove it. It might be confusion caused by the docstring of
>> latin-3-prefix, which says
>> 
>>   dot above  |   / .  | /g -> ġ   .o -> ġ
>> 
>> when itʼs actually .g -> ġ , with no mapping for 'o' at all.
>
> I think ~o is indeed a mistake, and should be deleted.  It makes no
> sense.

OK. Iʼve pushed the Romanian and latin-3-prefix doc changes as
3018a4779e , and the latin-pre mapping changes as 80c3b9aeba.

Thanks

Robert