RE: regex in csv2xml

I would do something like this:

<xsl:variable name="regex1">".*?"</xsl:variable>
<xsl:variable name="s1" as="xs:string*">
  <xsl:analyze-string select="$in" regex="{$regex1}">
    <xsl:matching-substring>
      <xsl:sequence select="replace(., '\n', &pua1;"/>
    </xsl:matching-substring>
    <xsl:non-matching-substring>
      <xsl:sequence select="."/>
    </xsl:non-matching-substring>
  </xsl:analyze-string>
</xsl:variable>

xsl:variable s2 select="string-join($s1, '')"

for each select tokenize($s2, '\n')
  for each select tokenize(., ',')
    replace(&pua1, '\n')

That is: first take the total string and identify substrings in quotes. The
fact that this treats "He said ""don't""" as three strings ("He said",
"don't", "") doesn't matter. Replace a newline appearing between quotes by a
private-use-area character (or any other 'spare' character). Then put the
strings back together again.

Now take the reassembled string and split it first at newlines, then at
commas, and within each identified token, convert the private character back
to a newline.

Michael Kay
http://www.saxonica.com/ 


> -----Original Message-----
> From: Jesper Tverskov [mailto:jesper@xxxxxxxxxxx] 
> Sent: 27 March 2006 08:51
> To: Xsl-List@Lists. Mulberrytech. Com
> Subject:  regex in csv2xml
> 
> Hi list,
> 
> I am trying to make a csv2xml XSLT 2.0 stylesheet using the 
> Excel csv format
> as example:
> If delimiter, newline or quotes are part of data the data is 
> quoted, quotes
> are doubled.
> 
> My last problem is that the newline character can be part of 
> data. I would
> like to detect thise newline characters and replace them 
> temporarily with
> some unique code.
> But have can I detect them in the first place?
> 
> Look at the sample below, we have 3 records and 3 fields:
> 
> 34,"""yes"", I said",46
> 25,"I said:
> ""Hello"", and I added: ""nice day, stranger""
> and, ""look at the sun"" , and: 
> ""bye for now.""",33
> 47,,35
> 
> Line 1 and 6 are records. We have an empty field in line 6.
> But line 2, 3, 4, 5 are one record with three linefeeds and 
> several commas
> as part of data.
> 
> How can I detect with a regex, that the linefeeds at the end 
> of line 2, 3
> and 4 are part of data?
> As I see it line 2 and 5 are the easy part, they will always 
> have an uneven
> number of quotes.
> But the linefeeds in line 3 and 4 can only be detected as 
> part of data if we
> compare all the lines being part of a record?
> 
> Best regards,
> Jesper Tverskov