Arrange code blocks sections based on dates and months

Question

I have a file with repeated contents like this;

<item>
    <date>August 24, 2021</date>
    <p>Text</p>
</item>
<item>
    <date>February 11, 2020</date>
    <p>more text</p>
</item>
<item>
    <date>July 20, 2021</date>
    <p>some text</p>
</item>

I was wishing to get something where the whole item sections will get arranged by date, where the first section item is the latest date and the last section item is of oldest date, something like this;

<item>
    <date>August 24, 2021</date>
    <p>Text</p>
</item>
<item>
    <date>July 20, 2021</date>
    <p>some text</p>
</item>
<item>
    <date>February 11, 2020</date>
    <p>more text</p>
</item>

Are there any possibilities of doing it with sed or awk?

You seem to be working with structured data (XML). In that case, it is usually a bad idea to use line-oriented tools like awk or sed, and dedicated parsers such as xmlstarlet should be used. Please indicate what you already tried and where you faced problems, so that contributors can help you find the solution to a specific question along the path. — AdminBee, Aug 26 '21 at 10:17
@AdminBee I haven't tried any codes yet, I couldn't figure out how to do it, maybe in a for loop then use sed and then do append, I dont know. The xmlstarlet, I will look that up, thanks — atheros, Aug 26 '21 at 10:21
The right answer will not involve a shell loop calling sed, see why-is-using-a-shell-loop-to-process-text-considered-bad-practice for some of the reasons why. — Ed Morton, Aug 26 '21 at 12:11
Somewhat related: Sorting an XML file in UNIX with a Bash script? — steeldriver, Aug 26 '21 at 13:08

Ed Morton · Accepted Answer · 2021-08-26T12:44:27.610

Hopefully someone will help you with an answer that uses an XML-aware tool but if not and assuming your input really does look like the sample you provided - using GNU awk for sorted_in:

$ cat tst.awk
BEGIN { RS=""; ORS="\n\n"; FS="</?date>" }
{
    split($2,d,/[, ]+/)
    mthAbbr = substr(d[1],1,3)
    mthNr = ( index( "JanFebMarAprMayJunJulAugSepOcNovDec", mthAbbr ) + 2 ) / 3
    date = sprintf("%04d%02d%02d",d[3], mthNr, d[2])
    items[date] = $0
}
END {
    PROCINFO["sorted_in"] = "@ind_num_desc"
    for ( date in items ) {
        print items[date]
    }
}

$ awk -f tst.awk file
<item>
    <date>August 24, 2021</date>
    <p>Text</p>
</item>
<item>
    <date>July 20, 2021</date>
    <p>some text</p>
</item>
<item>
    <date>February 11, 2020</date>
    <p>more text</p>
</item>

or using any awk plus sort and cut:

$ cat tst.awk
BEGIN { RS=""; FS="\n"; OFS="\t" }
{
    split($2,d,/[<>, ]+/)
    mthAbbr = substr(d[3],1,3)
    mthNr = ( index( "JanFebMarAprMayJunJulAugSepOcNovDec", mthAbbr ) + 2 ) / 3
    date = sprintf("%04d%02d%02d",d[5], mthNr, d[4])
for (i=1; i&lt;=NF; i++) {
    print date, NR, i, $i
}
print date, NR, i, &quot;&quot;

}

$ awk -f tst.awk file | sort -k1,1rn -k2,3n | cut -f4-
<item>
    <date>August 24, 2021</date>
    <p>Text</p>
</item>
<item>
    <date>July 20, 2021</date>
    <p>some text</p>
</item>
<item>
    <date>February 11, 2020</date>
    <p>more text</p>
</item>

The 2nd one will be a better choice if your input file is huge since it doesn't require awk to hold the whole input file in memory before printing it. It works by decorating the input lines to add the date for each item followed by the current record (item) number followed by the current line number within that item so that sort can then sort by date but retain the original input order even for duplicate dates, and then cut just removes the decorations that the first awk added to facilitate sorting. Here's what the output from the first 2 steps looks like so you can see what they do:

$ awk -f tst.awk file
20210824        1       1       <item>
20210824        1       2           <date>August 24, 2021</date>
20210824        1       3           <p>Text</p>
20210824        1       4       </item>
20210824        1       5
20200211        2       1       <item>
20200211        2       2           <date>February 11, 2020</date>
20200211        2       3           <p>more text</p>
20200211        2       4       </item>
20200211        2       5
20210720        3       1       <item>
20210720        3       2           <date>July 20, 2021</date>
20210720        3       3           <p>some text</p>
20210720        3       4       </item>
20210720        3       5

$ awk -f tst.awk file | sort -k1,1rn -k2,3n
20210824        1       1       <item>
20210824        1       2           <date>August 24, 2021</date>
20210824        1       3           <p>Text</p>
20210824        1       4       </item>
20210824        1       5
20210720        3       1       <item>
20210720        3       2           <date>July 20, 2021</date>
20210720        3       3           <p>some text</p>
20210720        3       4       </item>
20210720        3       5
20200211        2       1       <item>
20200211        2       2           <date>February 11, 2020</date>
20200211        2       3           <p>more text</p>
20200211        2       4       </item>
20200211        2       5

This absolutely works like a charm! I'm going with option 1. — atheros, Aug 26 '21 at 14:46

score 4 · Answer 2 · answered Aug 26 '21 at 13:29

Assuming that the <item>s are part of some container <foo>

$ cat file.xml
<foo>
<item>
    <date>August 24, 2021</date>
    <p>Text</p>
</item>
<item>
    <date>February 11, 2020</date>
    <p>more text</p>
</item>
<item>
    <date>July 20, 2021</date>
    <p>some text</p>
</item>
</foo>

then using xq from the yq project to leverage jq's time parsing and sorting capabilities:

$ xq -x  '.foo.item |= sort_by(now - (.date | strptime("%B %d, %Y") | mktime))' file.xml
<foo>
  <item>
    <date>August 24, 2021</date>
    <p>Text</p>
  </item>
  <item>
    <date>July 20, 2021</date>
    <p>some text</p>
  </item>
  <item>
    <date>February 11, 2020</date>
    <p>more text</p>
  </item>
</foo>

You could flip the sign of the mktime result to simplify it a bit: .[].item |= sort_by(.date | strptime("%b %d, %Y") | -mktime). — Kusalananda, Aug 26 '21 at 14:23
I'm going with Ed Morton's answer for now since it doesn't need extra package and absolutely works like a charm! — atheros, Aug 26 '21 at 14:49

Arrange code blocks sections based on dates and months

2 Answers2