How to initialize an array with elements with whitespace from a command output

Question

The following script:

DYN_HOSTS_START_ARRAY=($(grep -E "STARTING HOST" sample.log | cut -d' ' -f 1,2))
for ((i=0; i< ${#DYN_HOSTS_START_ARRAY[@]}; i++))
do
    echo "$i:  start: "${DYN_HOSTS_START_ARRAY[$i]}""
done

Using the following sample.log file:

2019-11-11 19:05:55,823 DEBUG  STARTING HOST 46
2019-11-11 19:05:55,831 DEBUG  STARTING HOST 703
2019-11-11 19:05:55,837 DEBUG  STARTING HOST 505
2019-11-11 19:05:55,858 DEBUG  STARTING HOST 93
2019-11-11 19:05:55,859 DEBUG  STARTING HOST 486
2019-11-11 19:05:55,861 DEBUG  STARTING HOST 72
2019-11-11 19:05:55,879 DEBUG STARTING HOST 855
2019-11-11 19:05:55,913 DEBUG  STARTING HOST 560
2019-11-11 19:05:56,067 DEBUG STARTING HOST 199

Yields the following undesired output:

0:  start: 2019-11-11
1:  start: 19:05:55,823
2:  start: 2019-11-11
3:  start: 19:05:55,831
4:  start: 2019-11-11
5:  start: 19:05:55,837
6:  start: 2019-11-11
7:  start: 19:05:55,858
8:  start: 2019-11-11
9:  start: 19:05:55,859
10:  start: 2019-11-11
11:  start: 19:05:55,861
12:  start: 2019-11-11
13:  start: 19:05:55,879
14:  start: 2019-11-11
15:  start: 19:05:55,913
16:  start: 2019-11-11
17:  start: 19:05:56,067

The desired output should only have 9 elements (instead of 18), each element containing both the date and the time, separated by the original space.

How do I fix my script, while retaining the array initialization, to 9 elements only, to accomplish that?

Answer 1

You could use mapfile -t and read the your data from a process substitution as @Kusalananda pointed out.

mapfile -t dyn_hosts_start_array < <(grep 'STARTING HOST' sample.log | cut -d' ' -f 1,2)    
for i in "${!dyn_hosts_start_array[@]}"; do
  printf '%s:  start: %s\n' "$i" "${dyn_hosts_start_array[i]}"
done

---

Or you could change the IFS variable which is used for word splitting to a newline character (default value is space character, tab and newline) and change it back to its original value afterwards.

oldifs=$IFS
IFS=$'\n'
dyn_hosts_start_array=( $(grep 'STARTING HOST' sample.log | cut -d' ' -f 1,2) )
IFS=$oldifs    
for i in "${!dyn_hosts_start_array[@]}"; do
  printf '%s:  start: %s\n' "$i" "${dyn_hosts_start_array[i]}"
done

---

Output (both variants):

0:  start: 2019-11-11 19:05:55,823
1:  start: 2019-11-11 19:05:55,831
2:  start: 2019-11-11 19:05:55,837
3:  start: 2019-11-11 19:05:55,858
4:  start: 2019-11-11 19:05:55,859
5:  start: 2019-11-11 19:05:55,861
6:  start: 2019-11-11 19:05:55,879
7:  start: 2019-11-11 19:05:55,913
8:  start: 2019-11-11 19:05:56,067

Related:

• What is the meaning of IFS=$'\n' in bash scripting?

Answer 2

awk '{print NR-1 ": ", "start:", $1, $2;}' sample.log

with the input from the question prints

0:  start: 2019-11-11 19:05:55,823
1:  start: 2019-11-11 19:05:55,831
2:  start: 2019-11-11 19:05:55,837
3:  start: 2019-11-11 19:05:55,858
4:  start: 2019-11-11 19:05:55,859
5:  start: 2019-11-11 19:05:55,861
6:  start: 2019-11-11 19:05:55,879
7:  start: 2019-11-11 19:05:55,913
8:  start: 2019-11-11 19:05:56,067

To have more control over the output formatting you could also use printf like this

awk '{printf "%d:  start: %s %s\n", NR-1, $1, $2;}' sample.log

Command substitution has some pitfalls with whitespace or special characters. Apparently, the two fields from the input are saved in separate array elements. This script combines them based on your original script.

DYN_HOSTS_START_ARRAY=($(grep -E "STARTING HOST" sample.log | cut -d' ' -f 1,2))
for ((i=0; i< ((${#DYN_HOSTS_START_ARRAY[@]} / 2)); i++))
do
    echo "$i:  start: ${DYN_HOSTS_START_ARRAY[((2 * $i))]} ${DYN_HOSTS_START_ARRAY[((2 * $i + 1))]}"
done