how to replace character in html attribute value (shell / bash)? - bash

Question

Sorry for the stupid question, but I have been stuck all afternoon with this simple problem. So I have a sample text file containing:
<product productId="123456" description="good apple, very green" publicPriceTTC="5,07" brand-id="152" />
<product productId="123457" description="fresh orange, very juicy" publicPriceTTC="12,47" brand-id="153" />
<product productId="123458" description="big banana, very yellow" publicPriceTTC="5,07" brand-id="154" />
And I'd like to modify this file into:
<product productId="123456" description="good apple, very green" publicPriceTTC="5.07" brand-id="152" />
<product productId="123457" description="fresh orange, very juicy" publicPriceTTC="12.47" brand-id="153" />
<product productId="123458" description="big banana, very yellow" publicPriceTTC="5.07" brand-id="154" />
Basically, I need to replace the "," (comma) by a "." (point) in all values of "publicPriceTTC". The trick here is that other attributes might have commas in their values ("description" in this example). I guess sed or awk can do that but I was unable to achieve it.
Can someone help me? Thank you very much for any help.

Answer 1

If you search for a comma to replace with a point, you will be doing a very coarse search/replace. Try something more especific. With sed, assume your input file is called xml:
sed -E 's/(publicPriceTTC="[0-9]+),([0-9]+")/\1.\2/' xml
You probably know that sed has the command s/<what you search>/<replacement>. We use that.
The -E option triggers the use of extended regular expressions. With that the s expression matches the whole tag + "=" + number within quotes, and uses the parenthesis to use the bit within them to be part of the substitution. \1 stands for the first bit between parenthesis block; \2 for the second.
You could of course make the search more robust to cope with whitespace between the tag and the equal sign and so on.

Answer 2

An awk solution to this might be:
awk '/<product/{for(i=1;i<=NF;i++){if($i~/^publicPriceTTC="/)sub(/,/,".",$i)}}1' file.xml
This steps through every whitespace-separated "field" on every <product>, looking for "words" that begin with the attribute you're trying to modify. If found, the entire attribute has its commas replaced with periods.
A simpler awk solution to emulate what others are doing with sed would be nice, except that awk does not support parenthesized subexpressions (i.e. \1 in your replacement string). Gawk supports them in the gensub() function, so the following might suffice:
gawk '{print gensub(/(publicPriceTTC="[0-9]+),/,"\\1.","g")}' file.xml
But ... you are solving the wrong problem here. Tools like sed and awk, which process files based on regular expressions, are not XML parsers. Either Javier's sed solution or my awk solutions could garble things accidentally, or miss certain things that are in perfectly valid XML files. Regex cannot be used to parse XML safely.
I recommend that you look into using python or perl or ruby or php or some other language with native XML support.
For example, turning your input into actual XML like this:
<p>
<product productId="123456" description="good apple, very green" publicPriceTTC="5,07" brand-id="152" />
<product productId="123457" description="fresh orange, very juicy" publicPriceTTC="12,47" brand-id="153" />
<product productId="123458" description="big banana, very yellow" publicPriceTTC="5,07" brand-id="154" />
</p>
We could run a PHP one-liner:
php -r '$x=new SimpleXMLElement(file_get_contents("file.xml")); foreach($x->product as $p) { $p["publicPriceTTC"]=str_replace(",",".",$p["publicPriceTTC"]); } print $x->asXML();'
Or split out for easier reading (and commenting):
<?php
// Read an XML file into an object
$x=new SimpleXMLElement(file_get_contents("file.xml"));
// Step through the object, fixing attributes as we find them
foreach($x->product as $p) {
$p["publicPriceTTC"] = str_replace(",",".",$p["publicPriceTTC"]);
}
// Print the result
print $x->asXML();

Answer 3

This will work on GNU
sed 's/\(publicPriceTTC="[0-9]*\),/\1./' fileName

Answer 4

Here using sub in awk is enough.
awk '{sub(/,/,".",$7)}1' file