Question

I am working on a bash script using sed and cut that will take times input in various ways and output them in a specific format. Here is an example line:

timeinhour=$(cut -d" " -f2<<<"$line" | sed 's/p/ /'  | sed 's/a/ /' | sed 's/am/ /' | sed 's/pm/ /' | sed 's/AM/ /' | sed 's/PM/ /' )

As you can see I am just removing any trailing am or pm from a time entry that might be formatted in various ways leaving only the numbers.

So I want this line to just spit out the hour of the day (timeinhour), ie “1000AM” = “10” as does “10a” and “10am.”

The problem I am running into is the varying lengths of the time entries. If I tell sed or cut to remove the last two characters “1000” will correctly output the hour I need: “10,” but using it on one that is already “10” obviously results in a blank output.

I have been experimenting with a line like this

sed 's/\(.*\)../\1/'

If anyone has any advice, I would appreciate it.

For example, this input:

1p
1032AM
419pm
1202a

would produce:

1
10
4
12

Answer 1

sed 's/[^0-9]//g;s/^[0-9]\{1,2\}$/&00/;s/^\(.*\)..$/\1/'

the steps

1p    -> 1    ->  100 -> 1
10a   -> 10   -> 1000 -> 10
419pm -> 419  ->  419 -> 4
1202a -> 1202 -> 1202 -> 12

• delete what is not number
• expand 1 or 2 digit (hours) into 4 digit HHmm
• ignore last two charactes (minutes)