I am wondering if there is a way to search for all the files from a certain directory including subdirectories using a find command on AIX 6.x, before calling an external command (e.g. hlcat) to display/convert them into a readable format, which can then be piped through a grep command to find a pattern instead of using loops in the shell?

e.g. find . -type f -name “*.hl7” -exec hlcat {} | grep -l “pattern” \;

The above command would not work and I have to use a while loop to display the content and search for the pattern as follows:

find . -type f -name “*.hl7” -print | while read file; do
hlcat $file | grep -l “pattern”;
done

At the same time, these HL7 files have been renamed with round brackets which prevent them from being open without having to include double quotes around the file name.

e.g. hlcat (patient) filename.hl7 will fail to open.
        hlcat “(patient) filename.hl7” will work.

In short, I am looking for a clean concise one-liner approach within the find command and view and search their content these HL7 files with round bracket names.

Many thanks, George

P.S. HL7 raw data is made up of one continuous line and is not readable unless it is converted into a workable reading format using tools such as hlcat. in

CodePudding user response：

1. Finding the HL7 files that contain pattern:

• with bash:

#!/bin/bash

find . -type f -name '*.hl7' -exec printf '%s\0' {}   |
while IFS='' read -r -d '' filepath
do
    hlcat "$filepath" | egrep -q 'pattern' && printf '%s\n' "$filepath"
done

^{note: AIX find doesn't have the -print0 option}

• or with a script inside find:

find . -type f -name '*.hl7' -exec sh -c '
    for f; do
        hlcat "$f" | egrep -q pattern && printf "%s\n" "$f";
    done
' _ {}  

^{note: pattern has to be correctly inserted in the script (will probably result in something less readable)}

2. Finding pattern in the HL7 files:

• as @Philippe suggested:

find . -type f -name '*.hl7' -exec hlcat {} \; | grep 'pattern'

• or if hlcat supports multiple files as argument:

find . -type f -name '*.hl7' -exec hlcat {}   | grep 'pattern'

Update: Replacing hlcat with cat to show you how it works

Consider the following files with simple content:

• filename: (Barry) fileX.hl7
  content: Barry

• filename: (John) fileY.hl7
  content: John

• filename: (Jolene) fileZ.hl7
  content: Jolene

The command:

find . -type f -name '*.hl7' -exec cat {} \; | egrep -l 'Barry|Jolene' 

Outputs:

(standard input)

That means that egrep found the pattern... in its standard input! That's all it can do, and your awk script will do something equivalent.

While the commands:

find . -type f -name '*.hl7' -exec sh -c '
    for f; do
        cat "$f" | egrep -q "Barry|Jolene" && printf "%s\n" "$f"; 
    done
' _ {}  

and

#!/bin/bash
find . -type f -name '*.hl7' -exec printf '%s\0' {}   |
while IFS='' read -r -d '' filepath
do
    cat "$filepath" | egrep -q 'Barry|Jolene' && printf '%s\n' "$filepath"
done

both output:

./(Jolene) fileZ.hl7
./(Barry) fileX.hl7

Which are the files that contain the pattern.

Now if you're fine without a robust handling of filenames, then your example with the while loop could work, provided that you add double quotes when expanding the filename:

find . -type f -name '*.hl7' -print | while read file; do
  cat "$file" | egrep -q 'Barry|Jolene' && printf '%s\n' "$file"
done

outputs:

./(Jolene) fileZ.hl7
./(Barry) fileX.hl7

But what's the point of not creating bullet-proof code when you can?

CodePudding user response：

A solution inspired of your awk example:

find . -type f -name '*.hl7' -exec awk -v regexp='^(Barry|Jolene)$' -F '|'  '
    FNR == 1 { if(found) print filename; found = 0; filename = FILENAME }
    $1 == "PID" && $5 ~ regexp { found = 1 }
    END { if (found) print filename }
' {}