Reading Command Output into Arrays with mapfile

1. mapfile reads lines into an indexed array

printf '%s\n' api worker cache > services.txt

mapfile -t services < services.txt

printf 'count=%d\n' "${#services[@]}"
printf '<%s>\n' "${services[@]}"

The -t option removes each line delimiter from the stored element. Without it, newline characters are retained.

2. readarray is a synonym

readarray -t services < services.txt
printf '%s\n' "${services[@]}"

Use whichever name your team finds clearer; both refer to the Bash builtin.

3. Process substitution connects command output without a pipeline subshell

mapfile -t services < <(
  printf '%s\n' "api gateway" worker cache
)

printf 'count=%d\n' "${#services[@]}"
printf '<%s>\n' "${services[@]}"

Each output line becomes one element, so embedded spaces remain intact.

4. Do not populate arrays with unquoted command substitution

# Fragile:
# services=( $(some_command) )

# Safe for line-oriented output:
mapfile -t services < <(some_command)

5. mapfile can use a custom delimiter

printf 'api\0worker\0cache\0' > services.bin

mapfile -d '' -t services < services.bin

printf '<%s>\n' "${services[@]}"

A NUL delimiter is useful for filename-safe records when paired with tools that produce NUL-separated output.

6. Read arbitrary filenames into an array

mapfile -d '' -t files < <(
  find ./artifacts -type f -print0
)

printf 'found=%d\n' "${#files[@]}"
printf 'path=%q\n' "${files[@]}"

This keeps spaces, tabs, and embedded newlines inside filenames from being misinterpreted as record separators.

7. -O appends starting at a chosen index

services=(api worker)

mapfile -t -O "${#services[@]}" services < <(
  printf '%s\n' cache metrics
)

printf '%s\n' "${services[@]}"

This is useful when building one array from several finite sources.

8. -s and -n can skip and limit input

printf '%s\n' a b c d e > values.txt

mapfile -t -s 1 -n 3 selected < values.txt

printf '%s\n' "${selected[@]}"

Here one record is skipped and at most three records are stored.

9. mapfile loads data into memory

For large or unbounded streams, a while read loop is usually better because it processes records incrementally.

while IFS= read -r line; do
  process_record "$line"
done < huge-input.txt

10. Producer failure needs explicit thought

Process substitution makes data capture convenient, but failure of the producer inside < <(...) is not represented as simply as a normal pipeline status.

# If source generation must succeed before proceeding,
# consider a validated temporary file:
tmp=$(mktemp)
if generate_inventory >"$tmp"; then
  mapfile -t items <"$tmp"
  rm -f "$tmp"
else
  status=$?
  rm -f "$tmp"
  exit "$status"
fi

11. Hands-on lab: load service inventory

mkdir -p "$HOME/devops-academy/bash/chapter07/lesson04"
cd "$HOME/devops-academy/bash/chapter07/lesson04"

inventory() {
  printf '%s\n' \
    "api gateway" \
    "worker" \
    "cache" \
    "metrics"
}

mapfile -t services < <(inventory)

printf 'services=%d\n' "${#services[@]}"

for service in "${services[@]}"; do
  printf 'service=<%s>\n' "$service"
done

mapfile -t -O "${#services[@]}" services < <(
  printf '%s\n' scheduler
)

printf 'after append=%d\n' "${#services[@]}"

12. Knowledge check

13. Summary

mapfile converts finite record streams directly into Bash arrays. It avoids command-substitution splitting, supports custom delimiters, and pairs well with process substitution. Use streaming instead when data is large or one-pass processing is sufficient.

14. Further reading

• GNU Bash Reference Manual — mapfile.
• GNU Bash Reference Manual — Process Substitution.
• GNU findutils manual — -print0.
• ShellCheck documentation — command substitution and arrays.