使用 sed 或通过匹配多行来编辑 xml 文件

Question

I'd like to 'edit' some xml files, which may have similar sections multiple times in one file. I need add 2 possible missing lines (i call it a pair) inside each section.. ie check if the 'pair' exists or not, if it does not then add them.

eg below is the possible missing pair lines i'd like to add in.

<arg>--possibleMissedKey</arg>
<arg>possibleMissedValue</arg>

Below file has the pair, so i do not need to add them in but if any of the section did miss this pair, i'd like to add the pair in to the section.. Also the number of lines in each section is not predictable.

    <some-tag-section-not-interesting>
        some contents not interesting to me
    </some-tag-section-not-interesting>
    <some-tag-to-look-for>
        <some stuff - a> ..... </some stuff - a>
        <arg>--possibleMissedKey</arg>
        <arg>possibleMissedValue</arg>
        <something-else-not-interesting>blahblah</something-else-not-interesting>
    </some-tag-to-look-for>
    <some-tag-to-look-for>
        <some stuff - b>....</some stuff - b>
        <arg>--possibleMissedKey</arg>
        <arg>possibleMissedValue</arg>
        <something-else-not-interesting>blahblah</something-else-not-interesting>
    </some-tag-to-look-for>

so i've consider a few options, but each one i've a question for it:

• the first thing came to my mind is 'sed'. I am hoping to replace the ending tag </some-tag-to-look-for> with

        <arg>--possibleMissedKey</arg>
        <arg>possibleMissedValue</arg>
     </some-tag-to-look-for> 

ie essentially add it to the last part of the section. but i am not aware of whether I can 'pattern match on multiple lines' in sed . and I have not used so called 'hold space'. My experience with sed has been to check some string in the 'current single line'.

• Another option i was hoping to check, is to introduce a inside_a_section_flag , initial value is '0'. start reading the file, the moment I find a staring <some-tag-to-look-for> , I turn that inside_a_section_flag to '1', and once I reach the 'ending </some-tag-to-look-for> ', I do some possible changes, and turn it back to '0'. so this inside_a_section_flag , if it is 1, means I are inside of the section.... so I need to look for the 'pair', if I found the pair, I turn the inside_a_section_flag to be 1, meaning I do not need to add the pair and can get out of the current section.... but I am not aware of whether sed can also work with a variable flag, ie do conditional replacement/change based on a variable value .

• Should this be done by shell at all, instead should this be done by a python script instead?

Answer 1

This might work for you (GNU sed):

sed '/<some-tag-to-look-for>/{:a;n;/<arg>--possibleMissedKey<\/arg>/b;/<\/some-tag-to-look-for>/!{h;ba};x;s/\S.*/<arg>--possibleMissedKey<\/arg>/p;s//<arg>--possibleMissedValue<\/arg>/p;x}' file

Match on a line containing <some-tag-to-look-for> .

Loop through the following lines.

If a line containing <arg>--possibleMissedKey</arg> is encountered, bail out.

Otherwise, if the current line does not match </some-tag-to-look-for> , make a copy and repeat.

When the end tag is found, insert the required two lines using the copied line as a template (so as to retain indentation).