How do I strip accents from characters in XSL?

Question

I keep looking, but can't find an XSL function that is the equivalent of "normalize-space", for characters. That is, my content has accented UNICODE characters, which is great, but from that content, I'm creating a filename, where I don't want those accents.

So, is there something that I'm overlooking, or not googling properly, to easily process characters?

In the XML data:

<filename>gri_gonéwiththèw00mitc</filename>

In XSLT stylesheet:

<xsl:variable name="file">
    <xsl:value-of select="filename"/>
</xsl:variable>

<xsl:value-of select="$file"/>

results in "gri_gonéwiththèw00mitc"

where

<xsl:value-of select='replace( normalize-unicode( "$file", "NFKD" ), "[^\\p{ASCII}]", "" )'/>

results in nothing.

What I'm aiming for is gri_gonewiththew00mitc (no accents)

Am I using the syntax wrong?

Intrigante

Removing accents only works for a small subset of Unicode characters. As far as I know, there's no standard way of latinized transcription of characters. (That is to say, there's a different one for each language.) – Ostracon

Check my answer for the correct RegExp syntax. – Blameworthy

Answer 1

In XSLT/XPath 1.0 if you want to replace those accented characters with the unaccented counterpart, you could use translate() function.

But, that assumes your "accented UNICODE characters" aren't composed unicode characters. If that were the case, you would need to use XPath 2.0 normalize-unicode() function.

And, if the real goal is to have a valid URI, you should use encode-for-uri()

Update: Examples

translate('gri_gonéwiththèw00mitc','áàâäéèêëíìîïóòôöúùûü','aaaaeeeeiiiioooouuuu')

Result: gri_gonewiththew00mitc

encode-for-uri('gri_gonéwiththèw00mitc')

Result: gri_gon%C3%A9withth%C3%A8w00mitc

Correct expression provide suggest by @biziclop:

replace(normalize-unicode('gri_gonéwiththèw00mitc','NFKD'),'\P{ASCII}','')

Result: gri_gonewiththew00mitc

Note: In XPath 2.0, the correct character class negation is with a capital \P.

Blameworthy

translate() assumes that you list all the characters you want to replace. My guess is that OP wants to avoid this. Although I don't think it's possible in general. – Ostracon

@biziclop: There is a reason for my answer having only one link to encode-for-uri() function. – Blameworthy

@Alejandro On second thought, if you normalize your string to NFKD form and then throw away every non-basic ASCII (0-127) character (you can use a regexp replace for that), you will get an accent-free string. – Ostracon

Answer 2

So, contrary to my comment, you could try this:

replace( normalize-unicode( "öt hűtőházból kértünk színhúst", "NFKD" ), "[^\\p{ASCII}]", "" )

Although be warned that any characters which can't be decomposed and aren't basic ASCII (Norwegian ø or Icelandic Þ for example) will be completely deleted from the string, but that's probably okay with your requirements.

Ostracon

Good example. Do check my update for the correct RegExp character class negation syntax. – Blameworthy

Where is this RegExp syntax example? I need to replace a list of characters.... ɔ ɛ and a few that I can`t even enter here.... As is noted these composed characters are deleted from the string using the above replace(normalize-unicode()) example. thx – Venerable

Answer 3

The previously suggested ways contain unknownthe character class named 'ASCII'. In my experience, XPath 2.0 recognises the class 'BasicLatin', which should serve the same purpose as 'ASCII'.

replace(normalize-unicode('Lliç d'Am Oükl Úkřeč', 'NFKD'), '\P{IsBasicLatin}', '')

Iffy

Answer 4

The top voted answer does not work anymore XPath2.0, as mentioned by Yuri. The 'IsBasicLatin' is an appropriate substitution for ASCII

The following code works:

replace(normalize-unicode('çgri_gonéwiththèmitç','NFKD'),'\P{IsBasicLatin}','')

Acerbity