How to format a string date (with text and milliseconds) using AWK

Question

I'm working on an AWK script that parses millions of lines of text. Each line contains (among other things) a date & time on the form:

16-FEB-2008 14:17:59.994669

I need to convert this into the following form

20080216141759994669000

And I would like avoid translating the month from text into a numerical value manually if it's possible. In bash I can simply do the following command to get the desired result:

date -d "16-FEB-2008 14:17:59.994669" +"%Y%m%d%H%M%S%N"

I have tried invoking this command into AWK but I cannot figure out howto. I would like to know

1. Is it possible to achieve with AWK alone?
2. How can I use such a command in an AWK script file?

Thanks in advance

Answer 1

Converting month names to numbers in awk is easy, and so is the reformatting as long as you don't need the (additional) validation date does 'for free':

$ echo this 16-FEB-2008 14:17:59.994669 that \
> | awk '{ split($2,d,"-"); split($3,t,"[:.]"); 
    m=sprintf("%02d",index("JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC",d[2])/4+1);
    print $1,d[3] m d[1] t[1] t[2] t[3] t[4] "000",$4 }'
this 20080216141759994669000 that
$ # or can put the script in a file and use with awk -f
$ # or the whole thing in a shebang file like #!/bin/awk -f

This is not much longer than the code to run date and much more efficient for 'millions of lines'.

Answer 2

You can call an external command like this:

awk '{
         cmd="date -d \""$0"\" +%Y%m%d%H%M%S%N"
         cmd | getline ts
         print $0, ts
         # awk opened a pipe for the communication with 
         # the command. close that pipe to avoid running
         # out of file descriptors
         close(cmd)
     }' <<< '16-FEB-2008 14:17:59.994669'

Output:

16-FEB-2008 14:17:59.994669 20080216141759994669000

---

Thanks to dave_thompson_085's comment you can significantly improve the performance if you have date from GNU coreutils and gawk. GNU's date supports reading dates from stdin and gawk supports co-processes which allows to start a single instance of date in the background, write into it's stdin and read from stdout:

{
    cmd = "stdbuf -oL date -f /dev/stdin +%Y%m%d%H%M%S%N"
    print $0 |& cmd 
    cmd |& getline ts
    print $0, ts
}

Note that you need to use the stdbuf command in addition to force date to output the results line by line.