Safe Path Handling and Filename Edge Cases

1. Treat pathnames as opaque data

A pathname is one shell argument, not a sentence. Quoting prevents word splitting and wildcard expansion from changing it.

file="release candidate.tar.gz"
printf 'path=<%s>\n' "$file"
cp -- "$file" backup/

2. Spaces and tabs are legal filename characters

mkdir -p demo
printf 'hello\n' > "demo/api build.log"
printf 'world\n' > $'demo/tab\tname.log'

for file in demo/*.log; do
  printf 'file=%q\n' "$file"
done

A glob can safely produce complete path words. Problems occur when the resulting variable is later expanded without quotes.

3. Filenames can begin with a dash

printf 'payload\n' > ./-dangerous
cat -- ./-dangerous
rm -- ./-dangerous

Many utilities recognize -- as the end of options. Prefixing a relative path with ./ also prevents it from beginning with a dash.

4. Newline is not a safe universal filename delimiter

Unix filenames may contain newline characters, so line-based filename lists are not fully general.

printf 'x\n' > $'demo/line\nbreak.txt'

find demo -type f -print0 |
while IFS= read -r -d '' file; do
  printf 'file=%q\n' "$file"
done

5. Wildcards can be literal filename characters

printf 'data\n' > 'demo/[prod]*.conf'
literal='demo/[prod]*.conf'
cat -- "$literal"

Quoted expansion keeps [, *, and ? as literal data instead of shell patterns.

6. Relative paths depend on the current directory

script_dir=$(
  cd -- "$(dirname -- "${BASH_SOURCE[0]}")" >/dev/null 2>&1 &&
  pwd -P
) || exit 1

config="$script_dir/../config/app.conf"

Script-owned assets should usually be resolved relative to the script, not to whichever directory the caller happened to use.

7. Symlinks create multiple filesystem questions

if [[ -L $path ]]; then
  printf 'symbolic link\n'
fi

8. Path text is not filesystem identity

Two different strings may resolve to the same file through ., .., hard links, or symlinks. Conversely, naive string normalization can change meaning when symlinks are involved.

9. Store multiple paths in arrays

files=(
  "artifacts/api build.tar"
  "artifacts/worker.tar"
  "artifacts/cache*.tar"
)

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

A space-delimited scalar is not a reliable pathname collection.

10. Hands-on lab: hostile pathname inventory

mkdir -p "$HOME/devops-academy/bash/chapter09/lesson01/files"
cd "$HOME/devops-academy/bash/chapter09/lesson01"

printf 'a\n' > "files/space name.txt"
printf 'b\n' > "files/-leading-dash.txt"
printf 'c\n' > 'files/[brackets]*.txt'
printf 'd\n' > $'files/newline\nname.txt'

count=0
while IFS= read -r -d '' file; do
  ((count += 1))
  printf 'path=%q bytes=%s\n' "$file" "$(wc -c < "$file")"
done < <(find files -type f -print0)

printf 'count=%d\n' "$count"

11. Knowledge check

12. Summary

Safe filesystem scripting treats pathnames as opaque data. Quote expansions, use option terminators where supported, preserve filename boundaries with NUL when required, and distinguish path text from the filesystem object it resolves to.

13. Further reading

• GNU Bash Reference Manual — Quoting and Filename Expansion.
• GNU Coreutils manuals — cp, rm, basename, dirname.
• GNU findutils manual — -print0.
• POSIX pathname and utility syntax conventions.