Remove unmatched parentheses and unmatched brackets from XML content - xslt-1.0

Question

I saw questions related for Ruby, etc. I didn't see any answers that would cover this case.
I'm using XSLT to transform an XML document.
I'm forced to use XSLT 1.0, since the application that uses XSLT and XPATH is still on 1.0.
Would you use Javascript function embedded in XSL using regex?
I have xml code like this:
<document>
<content name="PROD_MAJ_CLS_CD" type="text" vse-streams="2" u="20" action="cluster" weight="1">2</content>
<content name="PART_DESC_SHORT" type="text" vse-streams="2" u="22" action="cluster" weight="1">SCREW-ROCKER</content>
</document>
The content attribute where name="PART_DESC_SHORT" can have parentheses and brackets in it
Thanks,
Paul

Answer 1

In your example, none of the parentheses match, and all are open parentheses; for that case, the task of removing unmatched parentheses and brackets amounts to removing all parentheses and brackets. This is most easily done with the translate function:
<xsl:value-of select="translate(.,'([','')"/>
If you actually need to retain matched braces and translate input of the form a(b[c]d(e into ab[c]de, then the simplest way to do it in XSLT is with a recursive named template which takes two parameters: an input string and a stack recording the material processed so far. (The stack can be just a sequence of strings delimited by a separator you choose, that you hope won't appear in the input. Since you show the input as coming from an attribute value, &#x9; will probably be a safe choice (it will have been normalized out of the input unless the data producer used a numeric character reference for it), but you can use |||never-use-this-string-in-a-part-description||| instead if you like.)
On the initial call, you pass just the input string, and let the stack default to empty.
On each call with non-empty input, the template takes one character off of the input stream and does the right thing with it:
for "(" and "[", push a new string onto the stack.
for "]" and ")" matching the first character of the top item on the stack, pop the stack twice, concatenate item 2, item 1 (the old top item), and the character "]" or ")", and push the concatenation onto the stack.
For "]" and ")" that don't match, the input is not well balanced, and you need to do something. (The code here drops the character.)
For any other character, append it to the top item on the stack.
So at any given time, every item on the stack except the bottom one begins with an open parenthesis or bracket which has not yet been matched. Every time we get a match, the parenthesized string is appended to the next item on the stack, thus reducing the stack size.
When the input string is exhausted and the stack has more than one item in it, the leading brace on the stack's top item needs to be stripped and the top item (minus the brace) appended to the next item.
Once the stack is down to one item, that item contains the string you wanted.
Here it is in XSLT:
<xsl:template name="paren-match-or-strip">
<xsl:param name="input"/>
<xsl:param name="stack"/>
<xsl:variable name="char"
select="substring($input,1,1)"/>
<xsl:variable name="stacktop"
select="substring-before($stack,$sep)"/>
<xsl:variable name="stackrest"
select="substring-after($stack,$sep)"/>
<xsl:choose>
<xsl:when test="not($input = '')">
<xsl:choose>
<xsl:when test="$char = '(' or $char = '['">
<!--* Push another potential-left-brace on the stack *-->
<xsl:call-template name="paren-match-or-strip">
<xsl:with-param name="input"
select="substring($input,2)"/>
<xsl:with-param name="stack"
select="concat(
$char,
$sep,
$stack
)"/>
</xsl:call-template>
</xsl:when>
<xsl:when test="($char = ']' and substring($stacktop,1,1) = '[')
or
($char = ')' and substring($stacktop,1,1) = '(')
">
<!--* Match the left brace at the top of the stack *-->
<xsl:variable name="stacktop2"
select="substring-before($stackrest,$sep)"/>
<xsl:variable name="stackrest2"
select="substring-after($stackrest,$sep)"/>
<xsl:call-template name="paren-match-or-strip">
<xsl:with-param name="input"
select="substring($input,2)"/>
<xsl:with-param name="stack"
select="concat(
$stacktop2,
$stacktop,
$char,
$sep,
$stackrest
)"/>
</xsl:call-template>
</xsl:when>
<xsl:when test="$char = ']' or $char = ')'">
<!--* Unmatched right brace, drop it silently *-->
<xsl:call-template name="paren-match-or-strip">
<xsl:with-param name="input"
select="substring($input,2)"/>
<xsl:with-param name="stack"
select="$stack"/>
</xsl:call-template>
</xsl:when>
<xsl:otherwise>
<xsl:call-template name="paren-match-or-strip">
<xsl:with-param name="input"
select="substring($input,2)"/>
<xsl:with-param name="stack"
select="concat(
$stacktop,
$char,
$sep,
$stackrest
)"/>
</xsl:call-template>
</xsl:otherwise>
</xsl:choose>
</xsl:when>
<xsl:when test="$input = '' and contains($stackrest,$sep)">
<!--* Input is exhausted and at least one unmatched
* parenthesis is on the stack.
*-->
<xsl:variable name="stacktop2"
select="substring-before($stackrest,$sep)"/>
<xsl:variable name="stackrest2"
select="substring-after($stackrest,$sep)"/>
<xsl:call-template name="paren-match-or-strip">
<xsl:with-param name="input"
select="$input"/>
<xsl:with-param name="stack"
select="concat(
$stacktop2,
substring($stacktop,2),
$sep,
$stackrest2
)"/>
</xsl:call-template>
</xsl:when>
<xsl:otherwise>
<!--* input is exhausted, stack only has one item *-->
<xsl:value-of select="substring-before($stack,$sep)"/>
</xsl:otherwise>
</xsl:choose>
</xsl:template>