Re: Formatted output with locale

[Vladimir Zhbanov <vzhbanov@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 4928 bytes --]

Hi Ludovic,

...

>> What does it output if you type
>>   (number->locale-string 10000.00 2 (make-locale LC_ALL "fr_FR.utf8"))
>>
>> that is with two zeros after the point?
>
> I get this:
>
> --8<---------------cut here---------------start------------->8---
> scheme@(guile-user)> (number->locale-string 10000.00 2 (make-locale LC_ALL
> "fr_FR.utf8"))
> $1 = "10 000,0"
> scheme@(guile-user)> (number->locale-string 10000.00 4  (make-locale LC_ALL
> "fr_FR.utf8"))
> $2 = "10 000,0"
> --8<---------------cut here---------------end--------------->8---
>
> … and that’s definitely a bug.
>
> Could you send it to bug-guile@gnu.org so we keep track of it?

I don't know how to do this, sorry. I see bugs are numbered
there. Is there a special form for bug submission? Or just will an
email do? And how to answer the mail already being there if I'm
not subscribed?

>
>>> That’s on GNU/Linux (glibc 2.24).
>>>
>>> Note that number formatting data comes from the C library.  What C
>>> library do you use?
>>
>> My glibc version is 2.19.
>
> It might be responsible for the incorrect thousand and fraction
> separators you observe in the de_DE output.  To dig further, you can
> check the low-level info provided by ‘nl_langinfo’:
>
> --8<---------------cut here---------------start------------->8---
> scheme@(guile-user)> ,use(ice-9 i18n)
> scheme@(guile-user)> (define l (make-locale LC_ALL "de_DE.utf8"))
> scheme@(guile-user)> (locale-digit-grouping l)
> $2 = (3 3 . #-1#)
> scheme@(guile-user)> (locale-decimal-point l)
> $3 = ","
> scheme@(guile-user)> (locale-thousands-separator l)
> $4 = "."
> --8<---------------cut here---------------end--------------->8---


My Scheme output is as follows:
--------------------------------8<--------------------------------
scheme@(guile-user)> ,use (ice-9 i18n)
scheme@(guile-user)> (define l (make-locale LC_ALL "ru_RU.UTF-8"))
scheme@(guile-user)> (locale-digit-grouping l)
$2 = (3 3 . #-1#)
scheme@(guile-user)> (locale-decimal-point l)
$3 = ","
scheme@(guile-user)> (locale-thousands-separator l)
$4 = "\xa0"
-------------------------------->8--------------------------------

(The last result is weird on my system though it's another story.)

I don't think all this somehow affects my output. I seem to have
found several bugs in number->locale-string().


1) Superfluous minus sign in output of zero:

--------------------------------8<--------------------------------
scheme@(guile-user)> (number->locale-string 0)
$5 = "0"
scheme@(guile-user)> (number->locale-string 0.0)
$6 = "-0,0"
-------------------------------->8--------------------------------

For this case a patch against the master branch is attached.


2) Wrong output for numbers less than 0.001

--------------------------------8<--------------------------------
scheme@(guile-user)> (number->locale-string 0.0009)
$7 = "9,0e-4"
scheme@(guile-user)> (number->locale-string 0.0009 2)
$8 = "9,0e"
scheme@(guile-user)> (number->locale-string 0.0009 1)
$9 = "9,0"
scheme@(guile-user)> (number->locale-string 0.0009 0)
$10 = "9"
-------------------------------->8--------------------------------

I don't really know how to proceed here. My naive approach is to
not modify the decimal part in such cases. The solution is in the
second patch.

3) Should output for the below cases differ?
--------------------------------8<--------------------------------
scheme@(guile-user)> (number->locale-string 10 2)
$2 = "10"
scheme@(guile-user)> (number->locale-string 10.0 2)
$3 = "10,0"
-------------------------------->8--------------------------------

4) Definitely a bug with different fraction-digits:
--------------------------------8<--------------------------------
scheme@(guile-user)> (number->locale-string 10.0 2)
$3 = "10,0"
scheme@(guile-user)> (number->locale-string 10.0 4)
$4 = "10,0"
-------------------------------->8--------------------------------

I think, the decimal parts should be zero padded to get desired
result.

5) Eventually, I don't really understand why substring() is used
to cut decimals:

--------------------------------8<--------------------------------
scheme@(guile-user)> (number->locale-string 0.99 0)
$8 = "0"
scheme@(guile-user)> (number->locale-string 0.99 1)
$9 = "0,9"
scheme@(guile-user)> (number->locale-string 0.99 2)
$10 = "0,99"
-------------------------------->8--------------------------------

I would prefer some rounding here in order to have the result as
follows:

--Desired results---------------8<-------------------------Begin--
scheme@(guile-user)> (number->locale-string 0.99 0)
$8 = "1"
scheme@(guile-user)> (number->locale-string 0.99 1)
$9 = "1,0"
scheme@(guile-user)> (number->locale-string 0.99 2)
$10 = "0,99"
--Desired results--------------->8---------------------------End--

Regards,
  Vladimir

[-- Attachment #2: 0001-Fix-output-of-0.0-in-number-locale-string.patch --]
[-- Type: application/octet-stream, Size: 936 bytes --]

From 679acd3803b07e63a1d3a221d6b13eee969a745e Mon Sep 17 00:00:00 2001
From: Vladimir Zhbanov <vzhbanov@gmail.com>
Date: Wed, 23 Nov 2016 12:56:09 +0300
Subject: [PATCH 1/2] Fix output of 0.0 in number->locale-string().

---
 module/ice-9/i18n.scm | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/module/ice-9/i18n.scm b/module/ice-9/i18n.scm
index 1d12dd0..5e7e1cb 100644
--- a/module/ice-9/i18n.scm
+++ b/module/ice-9/i18n.scm
@@ -387,7 +387,7 @@ number of fractional digits to be displayed."
                                    (substring dec 0 fraction-digits)
                                    dec))))))
 
-    (let* ((external-repr (number->string (if (> number 0)
+    (let* ((external-repr (number->string (if (>= number 0)
                                               number
                                               (- number))))
            (int+dec   (string-split external-repr #\.))
-- 
2.1.4


[-- Attachment #3: 0002-Fix-number-locale-string-for-numbers-less-than-.001.patch --]
[-- Type: application/octet-stream, Size: 2091 bytes --]

From 3c2e718eff6ab1d8f529185896a4ea841122e0f0 Mon Sep 17 00:00:00 2001
From: Vladimir Zhbanov <vzhbanov@gmail.com>
Date: Wed, 23 Nov 2016 14:26:55 +0300
Subject: [PATCH 2/2] Fix number->locale-string() for numbers less than .001.

For such numbers, internally used number->string can return strings
containing exponent, such as, say "9.0e-4".  Just cutting of the decimal
part in this case is wrong, so the solution presented here is just to
use decimal part unchanged in such cases.
---
 module/ice-9/i18n.scm | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

diff --git a/module/ice-9/i18n.scm b/module/ice-9/i18n.scm
index 5e7e1cb..0c4d9f6 100644
--- a/module/ice-9/i18n.scm
+++ b/module/ice-9/i18n.scm
@@ -378,14 +378,17 @@ number of fractional digits to be displayed."
                 (else         "")))
          (decimal-part
           (lambda (dec)
-            (if (or (string=? dec "") (eq? 0 fraction-digits))
-                ""
-                (string-append (locale-decimal-point locale)
-                               (if (and (integer? fraction-digits)
-                                        (< fraction-digits
-                                           (string-length dec)))
-                                   (substring dec 0 fraction-digits)
-                                   dec))))))
+            (if (string-index dec #\-)
+                ;; For strings like "1.0e-4"
+                (string-append (locale-decimal-point locale) dec)
+                (if (or (string=? dec "") (eq? 0 fraction-digits))
+                 ""
+                 (string-append (locale-decimal-point locale)
+                                (if (and (integer? fraction-digits)
+                                         (< fraction-digits
+                                            (string-length dec)))
+                                    (substring dec 0 fraction-digits)
+                                    dec)))))))
 
     (let* ((external-repr (number->string (if (>= number 0)
                                               number
-- 
2.1.4