Currency symbol parsing code should ignore alternate symbols

[jtwang]

CLDR defines several formats for currency symbols (narrow, variant), Babel currently exposes the main symbol only. Eg. In es.xml, 'KMF' is 'main' symbol for currency KMF.

            <currency type="KMF">
                <displayName>franco comorense</displayName>
                <displayName count="one">franco comorense</displayName>
                <displayName count="other">francos comorenses</displayName>
                <symbol>KMF</symbol>
                <symbol alt="narrow">CF</symbol>
            </currency>

However, the code that selects and processes this element does not exclude the alternate forms, so in theory it's possible for an alternate symbol to be read and set as the symbol instead. This is does not actually manifest as a bug in Babel 2.3.x since CLDR 28 always defines the 'main' symbol element before the alt versions; however, this doesn't change the fact that the code isn't quite right. :)

Demonstration

Current CLDR + import-cldr script work fine:

$ grep -A6 currency.*KMF cldr/core-28/common/main/es.xml | grep symbol
                <symbol>KMF</symbol>
                <symbol alt="narrow">CF</symbol>
$ make clean-cldr
$ make import-cldr
$ python
>>> from babel.numbers import *
>>> get_currency_symbol('KMF', locale='es')
u'KMF'

Now switch the two symbol lines in es.xml and try again:

$ grep -A6 currency.*KMF cldr/core-28/common/main/es.xml | grep symbol
                <symbol alt="narrow">CF</symbol>
                <symbol>KMF</symbol>
$ make clean-cldr
$ make import-cldr
$ python
>>> from babel.numbers import *
>>> get_currency_symbol('KMF', locale='es')
u'CF'

A Possible Fix

Although it would be cool if we supported alt symbols. :)

diff --git a/scripts/import_cldr.py b/scripts/import_cldr.py
index 8de7043..c7d42d0 100755
--- a/scripts/import_cldr.py
+++ b/scripts/import_cldr.py
@@ -747,9 +747,9 @@ def parse_currency_names(data, tree):
             else:
                 currency_names[code] = text_type(name.text)
         # TODO: support choice patterns for currency symbol selection
-        symbol = elem.find('symbol')
-        if symbol is not None and 'draft' not in symbol.attrib and 'choice' not in symbol.attrib:
-            currency_symbols[code] = text_type(symbol.text)
+        for symbol in elem.findall('symbol'):
+            if symbol is not None and 'draft' not in symbol.attrib and 'choice' not in symbol.attrib and 'alt' not in symbol.attrib:
+                currency_symbols[code] = text_type(symbol.text)

[hemanthsavasere]

Hello,
Is anybody working on it, as I am eager to contribute and want to start working on this issue.

[benselme]

@kdeldycke Does your work on currencies relate to this in any way ?

[kdeldycke]

Thanks @benselme for the ping but I did not stumble onto that issue yet.

The thing is I'm still stuck on Babel 2.3 on my production servers. I'm waiting for #494 to be merged upstream to qualify the upcoming Babel 2.5. Maybe then I'll get the same issue.