如何计算awk中args的时差

Question

I have input in below format

0001580-160219044548744-oozie-oozi-W@:start:    -       -       -       -       OK      :start: 0       -       :START: 2016-02-24 13:34:46 GMT OK      2016-02-24 13:34:46 GMT
0001580-160219044548744-oozie-oozi-W@PrepareHDFS        -       -       -       -       OK      PrepareHDFS     0       -       fs      2016-02-24 13:34:46 GMT OK  2016-02-24 13:34:47 GMT

Using awk I am trying to print the required columns positioned in 7,12,16 Columns as below using the below command

cat input |  awk '/^000/ {printf "%-40s %-10s %-10s",$7,$12,$16}'

and the output is

:start:                                  05:35:56   05:35:56
PrepareHDFS                              05:35:56   05:35:57

My requirement is along with the above output I need time difference also. I tried below one inside a script

cat intput | awk '/^000/ {printf "%-40s %-10s %-10s",$7,$12,$16; 
T1=`date +%s -d $12`;T2=`date +%s -d $16`;
DIFF=`expr ${SEC2} - ${SEC1}`; print `date +%H:%M:%S -ud ${DIFF}` }'

But I am getting errors as invalid syntax. How can I achieve the time difference so that output should be

PrepareHDFS                              05:35:56   05:35:57  00:00:01
ScheduleStart                            05:35:57   05:36:11  00:00:14

EDIT:

For time difference, I have below script

TIME1=05:36:27
TIME2=05:36:51
SEC1=`date +%s -d ${TIME1}`
SEC2=`date +%s -d ${TIME2}`
DIFFSEC=`expr ${SEC2} - ${SEC1}`
echo `date +%H:%M:%S -ud @${DIFFSEC}`
00:00:24

Can I use this set of lines inside a function, and call that function from awk?

Answer 1

You can do it using awk , it's just a bit … unweildy:

awk -v cmd='date +%s -d ' -v cmd2='date +%H:%M:%S -d ' '/^000/ {
  cmd $12 | getline T1; close(cmd $12); 
  cmd $16 | getline T2; close(cmd $16); 
  cmd2 (T2 - T1) | getline T1; close(cmd2 (T2 - T1)); 
  printf "%-40s %-10s %-10s%-10s\n", $7, $12, $16, T1
}'

You can't use the shell's backtick process substitution in awk . awk has its own way of getting the output of a command - via getline and coprocesses . The syntax is roughly:

command | getline var-name
close(command)

Where command is a variable or string containing the command. cmd $12 is just the concatenation of cmd and $12 , so the command would become date +%s -d 13:34:46 , for example.

date interpretation of pure numbers as input is complex. It would be best to force it see the input number as a Unix timestamp, by using a leading @ :

awk -v cmd='date +%s -d ' -v cmd2='date +%H:%M:%S -d @' '/^000/ {
  cmd $12 | getline T1; close(cmd $12); 
  cmd $16 | getline T2; close(cmd $16); 
  cmd2 (T2 - T1) | getline T3; close(cmd2 (T2 - T1)); 
  printf "%-40s %-10s %-10s%-10s\n", $7, $12, $16, T3
}'