Regex fails with accented letters

[naoliv]

Using this example MapCSS rule:

*[name =~ /^[a-z ]+$/] {
        throwWarning: tr("name is all lowercase");
}

It should match both nodes from the attached example file, but it matches only the first one, with name=aa (failing in name=aá)

JOSM:

Build-Date:2016-02-21 00:01:27
Revision:9844
Is-Local-Build:true

Identification: JOSM/1.5 (9844 SVN pt_BR) Linux Debian GNU/Linux unstable (sid)
Memory Usage: 577 MB / 3641 MB (196 MB allocated, but free)
Java version: 1.8.0_72-internal-b15, Oracle Corporation, OpenJDK 64-Bit Server VM
VM arguments: [-Dawt.useSystemAAFontSettings=on]
Dataset consistency test: No problems found

Plugins:
- AddrInterpolation (31772)
- Create_grid_of_ways (31772)
- FastDraw (31895)
- FixAddresses (31772)
- OpeningHoursEditor (31772)
- PicLayer (31895)
- SimplifyArea (31895)
- apache-commons (31895)
- buildings_tools (31895)
- download_along (31772)
- editgpx (31772)
- ejml (31895)
- geotools (31895)
- graphview (31895)
- jts (31772)
- kendzi3d (1.0.189)
- kendzi3d-jogl (41)
- kendzi3d-resources (0.0.1)
- log4j (31895)
- measurement (31895)
- merge-overlap (31967)
- opendata (32071)
- pdfimport (32019)
- photo_geotagging (31895)
- poly (31772)
- reverter (32005)
- tagging-preset-tester (31895)
- todo (29154)
- turnrestrictions (31895)
- undelete (31895)
- utilsplugin2 (32018)

[wiktorn]

As far as I checked, [a-z] in Java doesn't cover national characters, nor is doing the same \w character class (​though last sentence in POSIX bracket expressions suggests that it should).

The class that covers all lowercase letters in Java is:
\p{javaLowerCase}

(See ​Pattern javadoc)

[naoliv]

Didn't know. Thanks for pointing this.

And how should I use it?

With *[name =~ /^[\p{javaLowerCase} ]+$/] there is a parsing exception.
With *[name =~ /^[\\p{javaLowerCase} ]+$/] it doesn't match anything.

[wiktorn]

My regex foo suggests something like:
*[name =~ /^(\p{javaLowerCase}| )+$/]

You can't use \p{...} character classes within [] AFAIK

[naoliv]

It fails with parsing error too.

[naoliv]

It seems that it also doesn't work with (?i):
For example:

*[name =~ /^(?i)fóo$/] won't match name=fÓo

[simon04]

Replying to naoliv:

*[name =~ /^(?i)fóo$/] won't match name=fÓo

You need to use [name =~ /^(?i)(?u)fóo$/], see ​https://docs.oracle.com/javase/7/docs/api/java/util/regex/Pattern.html#UNICODE_CASE

[simon04]

In 9857/josm:

fix #12549 - MapCSS: permit using character classes in regexp: \p{...}

[simon04]

Replying to naoliv:

It fails with parsing error too.

Fixed in r9857. Use [name =~ /^(?U)(\p{Lower})+$/], (?U) is compulsory, see ​https://docs.oracle.com/javase/7/docs/api/java/util/regex/Pattern.html#UNICODE_CHARACTER_CLASS

[naoliv]

Right. Thank you!