grobidOrg / grobid / 29514855442

Pull #1508

github

fix: do not prepend the underscore to training file names, only to xml:id

The leading underscore is an NCName constraint that only applies to the
xml:id values in the generated TEI. Split TextUtilities.sanitizeXmlId
into sanitizeFileName (character replacement only), used for the
training file names, and sanitizeXmlId which additionally prepends the
underscore when the name starts with a character not allowed as first
NCName character. Trainers resolving the feature file from the xml:id
strip the leading underscore, as TEIHeaderSaxParser.getPDFName already
does.

Signed-off-by: Luca Foppiano <luca@foppiano.org>

9841 of 25340 branches covered (38.84%)

Branch coverage included in aggregate %.

20524 of 46066 relevant lines covered (44.55%)

1.8 hits per line