Dear XSLT-List,
I'd be grateful for advice about a two-part <xsl:analyze-string> problem. I'm
post-processing messy OCR output, and the situation I'm trying to address
involves patterns and patterned errors that can be identified through regex
matching. Some of the patterns are traditional up-conversion (e.g., find a
certain pattern of digits and punctuation and wrap markup around it); some of
them are corrections (e.g., the digit "6" and the letter "b" are confused, but
a digit "6" adjacent to a letter is probably an error and should be corrected
automatically, while a digit "6" not adjacent to a letter probably isn't and
should be left alone).
1. The first part of my problem involves general program logic. I'm currently
using a strategy like the following:
<xsl:template match="text()">
<xsl:call-template name="editionLineNo">
<xsl:with-param name="current" select="."/>
</xsl:call-template>
</xsl:template>
<xsl:template name="editionLineNo">
<!-- 1. check for digits plus period, \d+\., edition line no -->
<xsl:param name="current"/>
<xsl:analyze-string select="$current" regex="(\d+)\.">
<xsl:matching-substring>
<editionLineNo>
<xsl:value-of select="regex-group(1)"/>
</editionLineNo>
</xsl:matching-substring>
<xsl:non-matching-substring>
<xsl:call-template name="msFolioNo">
<xsl:with-param name="current" select="$current"/>
</xsl:call-template>
</xsl:non-matching-substring>
</xsl:analyze-string>
</xsl:template>
That is, at the beginning I grab a pristine text node and look for a pattern.
If it's there, I'm done; if not, I pass the non-matching substring to the next
template to look for a different pattern. One template calls another, passing
the unmatched substrings, until the end, when I just output the text.
This works, but is it the best approach? Should I instead, for example, use a
single callable template and pass it both the haystack string and the needle
regex? My highest priorities are legibility and ease of development and
maintenance; efficiency of operation is less important. In case this is
important, the order in which the patterns are matched matters, at least in a
few instances. For example, digits followed by a period get one kind of markup
and digits not followed by a period get another, so I want to capture the first
type first and get them out of the way before looking for the second.
2. The second part of my problem involves a particular type of regex, one that
will, for example, identify a digit "6" that is adjacent to a letter and
replace it with a letter "b". The adjacent letter could precede or follow the
digit or both. If I make the preceding and following letter(s) optional in the
pattern, I've made both optional, and I'll erroneously catch an isolated digit
"6". If I use a disjunct pattern, it becomes harder to capture the pieces and
output the ones I want to retain with regex-group(). I suspect that this is a
common problem with a standard solution, but I haven't run into it before and
no single, elegant but legible regex leaps to mind. Is there one?
Thanks for any advice,,
David
djbpitt(_at_)gmail(_dot_)com
--~------------------------------------------------------------------
XSL-List info and archive: http://www.mulberrytech.com/xsl/xsl-list
To unsubscribe, go to: http://lists.mulberrytech.com/xsl-list/
or e-mail: <mailto:xsl-list-unsubscribe(_at_)lists(_dot_)mulberrytech(_dot_)com>
--~--