Related
x=$(find . -name "*.txt")
echo $x
if I run the above piece of code in Bash shell, what I get is a string containing several file names separated by blank, not a list.
Of course, I can further separate them by blank to get a list, but I'm sure there is a better way to do it.
So what is the best way to loop through the results of a find command?
TL;DR: If you're just here for the most correct answer, you probably want my personal preference (see the bottom of this post):
# execute `process` once for each file
find . -name '*.txt' -exec process {} \;
If you have time, read through the rest to see several different ways and the problems with most of them.
The full answer:
The best way depends on what you want to do, but here are a few options. As long as no file or folder in the subtree has whitespace in its name, you can just loop over the files:
for i in $x; do # Not recommended, will break on whitespace
process "$i"
done
Marginally better, cut out the temporary variable x:
for i in $(find -name \*.txt); do # Not recommended, will break on whitespace
process "$i"
done
It is much better to glob when you can. White-space safe, for files in the current directory:
for i in *.txt; do # Whitespace-safe but not recursive.
process "$i"
done
By enabling the globstar option, you can glob all matching files in this directory and all subdirectories:
# Make sure globstar is enabled
shopt -s globstar
for i in **/*.txt; do # Whitespace-safe and recursive
process "$i"
done
In some cases, e.g. if the file names are already in a file, you may need to use read:
# IFS= makes sure it doesn't trim leading and trailing whitespace
# -r prevents interpretation of \ escapes.
while IFS= read -r line; do # Whitespace-safe EXCEPT newlines
process "$line"
done < filename
read can be used safely in combination with find by setting the delimiter appropriately:
find . -name '*.txt' -print0 |
while IFS= read -r -d '' line; do
process "$line"
done
For more complex searches, you will probably want to use find, either with its -exec option or with -print0 | xargs -0:
# execute `process` once for each file
find . -name \*.txt -exec process {} \;
# execute `process` once with all the files as arguments*:
find . -name \*.txt -exec process {} +
# using xargs*
find . -name \*.txt -print0 | xargs -0 process
# using xargs with arguments after each filename (implies one run per filename)
find . -name \*.txt -print0 | xargs -0 -I{} process {} argument
find can also cd into each file's directory before running a command by using -execdir instead of -exec, and can be made interactive (prompt before running the command for each file) using -ok instead of -exec (or -okdir instead of -execdir).
*: Technically, both find and xargs (by default) will run the command with as many arguments as they can fit on the command line, as many times as it takes to get through all the files. In practice, unless you have a very large number of files it won't matter, and if you exceed the length but need them all on the same command line, you're SOL find a different way.
What ever you do, don't use a for loop:
# Don't do this
for file in $(find . -name "*.txt")
do
…code using "$file"
done
Three reasons:
For the for loop to even start, the find must run to completion.
If a file name has any whitespace (including space, tab or newline) in it, it will be treated as two separate names.
Although now unlikely, you can overrun your command line buffer. Imagine if your command line buffer holds 32KB, and your for loop returns 40KB of text. That last 8KB will be dropped right off your for loop and you'll never know it.
Always use a while read construct:
find . -name "*.txt" -print0 | while read -d $'\0' file
do
…code using "$file"
done
The loop will execute while the find command is executing. Plus, this command will work even if a file name is returned with whitespace in it. And, you won't overflow your command line buffer.
The -print0 will use the NULL as a file separator instead of a newline and the -d $'\0' will use NULL as the separator while reading.
find . -name "*.txt"|while read fname; do
echo "$fname"
done
Note: this method and the (second) method shown by bmargulies are safe to use with white space in the file/folder names.
In order to also have the - somewhat exotic - case of newlines in the file/folder names covered, you will have to resort to the -exec predicate of find like this:
find . -name '*.txt' -exec echo "{}" \;
The {} is the placeholder for the found item and the \; is used to terminate the -exec predicate.
And for the sake of completeness let me add another variant - you gotta love the *nix ways for their versatility:
find . -name '*.txt' -print0|xargs -0 -n 1 echo
This would separate the printed items with a \0 character that isn't allowed in any of the file systems in file or folder names, to my knowledge, and therefore should cover all bases. xargs picks them up one by one then ...
Filenames can include spaces and even control characters. Spaces are (default) delimiters for shell expansion in bash and as a result of that x=$(find . -name "*.txt") from the question is not recommended at all. If find gets a filename with spaces e.g. "the file.txt" you will get 2 separated strings for processing, if you process x in a loop. You can improve this by changing delimiter (bash IFS Variable) e.g. to \r\n, but filenames can include control characters - so this is not a (completely) safe method.
From my point of view, there are 2 recommended (and safe) patterns for processing files:
1. Use for loop & filename expansion:
for file in ./*.txt; do
[[ ! -e $file ]] && continue # continue, if file does not exist
# single filename is in $file
echo "$file"
# your code here
done
2. Use find-read-while & process substitution
while IFS= read -r -d '' file; do
# single filename is in $file
echo "$file"
# your code here
done < <(find . -name "*.txt" -print0)
Remarks
on Pattern 1:
bash returns the search pattern ("*.txt") if no matching file is found - so the extra line "continue, if file does not exist" is needed. see Bash Manual, Filename Expansion
shell option nullglob can be used to avoid this extra line.
"If the failglob shell option is set, and no matches are found, an error message is printed and the command is not executed." (from Bash Manual above)
shell option globstar: "If set, the pattern ‘**’ used in a filename expansion context will match all files and zero or more directories and subdirectories. If the pattern is followed by a ‘/’, only directories and subdirectories match." see Bash Manual, Shopt Builtin
other options for filename expansion: extglob, nocaseglob, dotglob & shell variable GLOBIGNORE
on Pattern 2:
filenames can contain blanks, tabs, spaces, newlines, ... to process filenames in a safe way, find with -print0 is used: filename is printed with all control characters & terminated with NUL. see also Gnu Findutils Manpage, Unsafe File Name Handling, safe File Name Handling, unusual characters in filenames. See David A. Wheeler below for detailed discussion of this topic.
There are some possible patterns to process find results in a while loop. Others (kevin, David W.) have shown how to do this using pipes:
files_found=1
find . -name "*.txt" -print0 |
while IFS= read -r -d '' file; do
# single filename in $file
echo "$file"
files_found=0 # not working example
# your code here
done
[[ $files_found -eq 0 ]] && echo "files found" || echo "no files found"
When you try this piece of code, you will see, that it does not work: files_found is always "true" & the code will always echo "no files found". Reason is: each command of a pipeline is executed in a separate subshell, so the changed variable inside the loop (separate subshell) does not change the variable in the main shell script. This is why I recommend using process substitution as the "better", more useful, more general pattern.See I set variables in a loop that's in a pipeline. Why do they disappear... (from Greg's Bash FAQ) for a detailed discussion on this topic.
Additional References & Sources:
Gnu Bash Manual, Pattern Matching
Filenames and Pathnames in Shell: How to do it Correctly, David A. Wheeler
Why you don't read lines with "for", Greg's Wiki
Why you shouldn't parse the output of ls(1), Greg's Wiki
Gnu Bash Manual, Process Substitution
(Updated to include #Socowi's execellent speed improvement)
With any $SHELL that supports it (dash/zsh/bash...):
find . -name "*.txt" -exec $SHELL -c '
for i in "$#" ; do
echo "$i"
done
' {} +
Done.
Original answer (shorter, but slower):
find . -name "*.txt" -exec $SHELL -c '
echo "$0"
' {} \;
If you can assume the file names don't contain newlines, you can read the output of find into a Bash array using the following command:
readarray -t x < <(find . -name '*.txt')
Note:
-t causes readarray to strip newlines.
It won't work if readarray is in a pipe, hence the process substitution.
readarray is available since Bash 4.
Bash 4.4 and up also supports the -d parameter for specifying the delimiter. Using the null character, instead of newline, to delimit the file names works also in the rare case that the file names contain newlines:
readarray -d '' x < <(find . -name '*.txt' -print0)
readarray can also be invoked as mapfile with the same options.
Reference: https://mywiki.wooledge.org/BashFAQ/005#Loading_lines_from_a_file_or_stream
# Doesn't handle whitespace
for x in `find . -name "*.txt" -print`; do
process_one $x
done
or
# Handles whitespace and newlines
find . -name "*.txt" -print0 | xargs -0 -n 1 process_one
I like to use find which is first assigned to variable and IFS switched to new line as follow:
FilesFound=$(find . -name "*.txt")
IFSbkp="$IFS"
IFS=$'\n'
counter=1;
for file in $FilesFound; do
echo "${counter}: ${file}"
let counter++;
done
IFS="$IFSbkp"
As commented by #Konrad Rudolph this will not work with "new lines" in file name. I still think it is handy as it covers most of the cases when you need to loop over command output.
As already posted on the top answer by Kevin, the best solution is to use a for loop with bash glob, but as bash glob is not recursive by default, this can be fixed by a bash recursive function:
#!/bin/bash
set -x
set -eu -o pipefail
all_files=();
function get_all_the_files()
{
directory="$1";
for item in "$directory"/* "$directory"/.[^.]*;
do
if [[ -d "$item" ]];
then
get_all_the_files "$item";
else
all_files+=("$item");
fi;
done;
}
get_all_the_files "/tmp";
for file_path in "${all_files[#]}"
do
printf 'My file is "%s"\n' "$file_path";
done;
Related questions:
Bash loop through directory including hidden file
Recursively list files from a given directory in Bash
ls command: how can I get a recursive full-path listing, one line per file?
List files recursively in Linux CLI with path relative to the current directory
Recursively List all directories and files
bash script, create array of all files in a directory
How can I creates array that contains the names of all the files in a folder?
How can I creates array that contains the names of all the files in a folder?
How to get the list of files in a directory in a shell script?
based on other answers and comment of #phk, using fd #3:
(which still allows to use stdin inside the loop)
while IFS= read -r f <&3; do
echo "$f"
done 3< <(find . -iname "*filename*")
You can put the filenames returned by find into an array like this:
array=()
while IFS= read -r -d ''; do
array+=("$REPLY")
done < <(find . -name '*.txt' -print0)
Now you can just loop through the array to access individual items and do whatever you want with them.
Note: It's white space safe.
You can store your find output in array if you wish to use the output later as:
array=($(find . -name "*.txt"))
Now to print the each element in new line, you can either use for loop iterating to all the elements of array, or you can use printf statement.
for i in ${array[#]};do echo $i; done
or
printf '%s\n' "${array[#]}"
You can also use:
for file in "`find . -name "*.txt"`"; do echo "$file"; done
This will print each filename in newline
To only print the find output in list form, you can use either of the following:
find . -name "*.txt" -print 2>/dev/null
or
find . -name "*.txt" -print | grep -v 'Permission denied'
This will remove error messages and only give the filename as output in new line.
If you wish to do something with the filenames, storing it in array is good, else there is no need to consume that space and you can directly print the output from find.
I think using this piece of code (piping the command after while done):
while read fname; do
echo "$fname"
done <<< "$(find . -name "*.txt")"
is better than this answer because while loop is executed in a subshell according to here, if you use this answer and variable changes cannot be seen after while loop if you want to modify variables inside the loop.
function loop_through(){
length_="$(find . -name '*.txt' | wc -l)"
length_="${length_#"${length_%%[![:space:]]*}"}"
length_="${length_%"${length_##*[![:space:]]}"}"
for i in {1..$length_}
do
x=$(find . -name '*.txt' | sort | head -$i | tail -1)
echo $x
done
}
To grab the length of the list of files for loop, I used the first command "wc -l".
That command is set to a variable.
Then, I need to remove the trailing white spaces from the variable so the for loop can read it.
find <path> -xdev -type f -name *.txt -exec ls -l {} \;
This will list the files and give details about attributes.
Another alternative is to not use bash, but call Python to do the heavy lifting. I recurred to this because bash solutions as my other answer were too slow.
With this solution, we build a bash array of files from inline Python script:
#!/bin/bash
set -eu -o pipefail
dsep=":" # directory_separator
base_directory=/tmp
all_files=()
all_files_string="$(python3 -c '#!/usr/bin/env python3
import os
import sys
dsep="'"$dsep"'"
base_directory="'"$base_directory"'"
def log(*args, **kwargs):
print(*args, file=sys.stderr, **kwargs)
def check_invalid_characther(file_path):
for thing in ("\\", "\n"):
if thing in file_path:
raise RuntimeError(f"It is not allowed {thing} on \"{file_path}\"!")
def absolute_path_to_relative(base_directory, file_path):
relative_path = os.path.commonprefix( [ base_directory, file_path ] )
relative_path = os.path.normpath( file_path.replace( relative_path, "" ) )
# if you use Windows Python, it accepts / instead of \\
# if you have \ on your files names, rename them or comment this
relative_path = relative_path.replace("\\", "/")
if relative_path.startswith( "/" ):
relative_path = relative_path[1:]
return relative_path
for directory, directories, files in os.walk(base_directory):
for file in files:
local_file_path = os.path.join(directory, file)
local_file_name = absolute_path_to_relative(base_directory, local_file_path)
log(f"local_file_name {local_file_name}.")
check_invalid_characther(local_file_name)
print(f"{base_directory}{dsep}{local_file_name}")
' | dos2unix)";
if [[ -n "$all_files_string" ]];
then
readarray -t temp <<< "$all_files_string";
all_files+=("${temp[#]}");
fi;
for item in "${all_files[#]}";
do
OLD_IFS="$IFS"; IFS="$dsep";
read -r base_directory local_file_name <<< "$item"; IFS="$OLD_IFS";
printf 'item "%s", base_directory "%s", local_file_name "%s".\n' \
"$item" \
"$base_directory" \
"$local_file_name";
done;
Related:
os.walk without hidden folders
How to do a recursive sub-folder search and return files in a list?
How to split a string into an array in Bash?
How about if you use grep instead of find?
ls | grep .txt$ > out.txt
Now you can read this file and the filenames are in the form of a list.
for subj in `cat dti_list.txt`; do
echo $subj
find . -type f -iname '*306.nii' -execdir bash -c 'rename.ul "$subj" DTI_MAIN_AP.nii *.nii' \+
done
I have some trouble with a small bash script, which adds the name instead of replacing when I use the rename.ul function.
Currently, the code adds DTI_MAIN_AP.nii in front of the old name.
My goal is to replace the name from the subj list and using the find to search up any directory with a *306.nii file, and then using execdir to execute the rename.ul function to rename the file from the dti_list.txt.
Any solution, or correction to get the code working, will be appreciated.
If you just want to rename the first file matching *306.nii in each directory to DTI_MAIN_AP.nii, that might look like:
find . -type f -iname '*306.nii' \
-execdir sh -c '[[ -e DTI_MAIN_AP.nii ]] || mv "$1" DTI_MAIN_AP.nii' _ {} +
If instead of matching on *306.nii you want to iterate over names from dti_list.txt, that might instead look like:
while IFS= read -r -d '' filename <&3; do
find . -type f -name "$filename" \
-execdir sh -c '[[ -e DTI_MAIN_AP.nii ]] || mv "$1" DTI_MAIN_AP.nii' _ {} +
done <dti_list.txt
References of note:
BashFAQ #1 (on reading files line-by-line)
Using Find
I am writing a script that wraps the find command to search for specific source file types under a given directory. A sample invocation would be :
./find_them.sh --java --flex --xml dir1
The above command would search for .java, .as and .xml files under dir1.
To do this manually I came up with the following find command :
find dir1 -type f -a \( -name "*.java" -o -name "*.as" -o -name "*.xml" \)
As I am doing this in a script where I want to be able specify different file sets to search for you end up with the following structure :
find_cmd_file_sets=$(decode_file_sets) # Assume this creates a string with the file sets e.g. -name "*.java" -o -name "*.as" etc
dirs=$(get_search_dirs) # assume this gives you the list of dirs to search, defaulting to the current directory
for dir in $dirs
do
find $dir -type f -a \( $find_cmd_file_sets \)
done
The above script doesn't behave as expected, you execute the script and the find command churns for a while before returning no results.
I'm certain the equivalents of decode_file_sets and get_search_dirs I've created are generating the correct results.
A simpler example if to execute the following directly in a bash shell
file_sets=' -name "*.java" -o -name "*.as" '
find dir -type f -a \( $file_sets \) # Returns no result
# Executing result of below command directly in the shell returns correct result
echo find dir -type f -a \\\( $file_sets \\\)
I don't understand why variable expansion in brackets of the find command would change the result. If it makes any difference I am using git-bash under Windows.
This is really frustrating. Any help would be much appreciated. Most importantly I would like to understand why the variable expansion of $file_sets is behaving as it is.
Hope this will work, Its tested on bash.
file_sets=' -name "*.java" -o -name "*.as" '
command=`echo "find $dir -type f -a \( $file_sets \)"`
eval $command
TLDR: Don't use quotes in find_cmd_file_sets variable and disable pathname expansion (set -f) before calling find.
When you have "special" character in a variable content and then you try to expand that variable without quotes than bash will surround each word with "special" character with single quotes, e.g.:
#!/usr/bin/env bash
set -x
VAR='abc "def"'
echo $VAR
The output is:
+ VAR='abc "def"'
+ echo abc '"def"'
abc "def"
As you can see, bash surrounded "def" with single quotes. In your case, the call to find command becomes:
find ... -name '"*.java"' ...
So it tries to find files which start with " and end with .java"
To prevent that behavior, the only thing you can do (which I'm aware of) is to use double quotes when expanding the variable, e.g.:
#!/usr/bin/env bash
set -x
VAR='abc "def"'
echo "$VAR"
The output is:
+ VAR='abc "def"'
+ echo 'abc "def"'
abc "def"
The only problem, as you probably noticed already, is that now the whole variable is in quotes and is treated as single argument. So this won't work in your find command.
The only option left is to not use quotes, neither in variable content nor when expanding the variable. But then, of course, you have a problem with pathname expansion:
#!/usr/bin/env bash
set -x
VAR='abc *.java'
echo $VAR
The output is:
+ VAR='abc *.java'
+ echo abc file1.java file2.java
abc file1.java file2.java
Fortunately you can disable pathname expansion using set -f:
#!/usr/bin/env bash
set -x
VAR='abc *.java'
set -f
echo $VAR
The output is:
+ VAR='abc *.java'
+ set -f
+ echo abc '*.java'
abc *.java
To sum up, the following should work:
#!/usr/bin/env bash
pattern='-name *.java'
dir="my_project"
set -f
find "$dir" -type f -a \( $pattern \)
bash arrays were introduced to allow this kind of nested quoting:
file_sets=( -name "*.java" -o -name "*.as" )
find dir -type f -a \( "${file_sets[#]}" \)
x=$(find . -name "*.txt")
echo $x
if I run the above piece of code in Bash shell, what I get is a string containing several file names separated by blank, not a list.
Of course, I can further separate them by blank to get a list, but I'm sure there is a better way to do it.
So what is the best way to loop through the results of a find command?
TL;DR: If you're just here for the most correct answer, you probably want my personal preference (see the bottom of this post):
# execute `process` once for each file
find . -name '*.txt' -exec process {} \;
If you have time, read through the rest to see several different ways and the problems with most of them.
The full answer:
The best way depends on what you want to do, but here are a few options. As long as no file or folder in the subtree has whitespace in its name, you can just loop over the files:
for i in $x; do # Not recommended, will break on whitespace
process "$i"
done
Marginally better, cut out the temporary variable x:
for i in $(find -name \*.txt); do # Not recommended, will break on whitespace
process "$i"
done
It is much better to glob when you can. White-space safe, for files in the current directory:
for i in *.txt; do # Whitespace-safe but not recursive.
process "$i"
done
By enabling the globstar option, you can glob all matching files in this directory and all subdirectories:
# Make sure globstar is enabled
shopt -s globstar
for i in **/*.txt; do # Whitespace-safe and recursive
process "$i"
done
In some cases, e.g. if the file names are already in a file, you may need to use read:
# IFS= makes sure it doesn't trim leading and trailing whitespace
# -r prevents interpretation of \ escapes.
while IFS= read -r line; do # Whitespace-safe EXCEPT newlines
process "$line"
done < filename
read can be used safely in combination with find by setting the delimiter appropriately:
find . -name '*.txt' -print0 |
while IFS= read -r -d '' line; do
process "$line"
done
For more complex searches, you will probably want to use find, either with its -exec option or with -print0 | xargs -0:
# execute `process` once for each file
find . -name \*.txt -exec process {} \;
# execute `process` once with all the files as arguments*:
find . -name \*.txt -exec process {} +
# using xargs*
find . -name \*.txt -print0 | xargs -0 process
# using xargs with arguments after each filename (implies one run per filename)
find . -name \*.txt -print0 | xargs -0 -I{} process {} argument
find can also cd into each file's directory before running a command by using -execdir instead of -exec, and can be made interactive (prompt before running the command for each file) using -ok instead of -exec (or -okdir instead of -execdir).
*: Technically, both find and xargs (by default) will run the command with as many arguments as they can fit on the command line, as many times as it takes to get through all the files. In practice, unless you have a very large number of files it won't matter, and if you exceed the length but need them all on the same command line, you're SOL find a different way.
What ever you do, don't use a for loop:
# Don't do this
for file in $(find . -name "*.txt")
do
…code using "$file"
done
Three reasons:
For the for loop to even start, the find must run to completion.
If a file name has any whitespace (including space, tab or newline) in it, it will be treated as two separate names.
Although now unlikely, you can overrun your command line buffer. Imagine if your command line buffer holds 32KB, and your for loop returns 40KB of text. That last 8KB will be dropped right off your for loop and you'll never know it.
Always use a while read construct:
find . -name "*.txt" -print0 | while read -d $'\0' file
do
…code using "$file"
done
The loop will execute while the find command is executing. Plus, this command will work even if a file name is returned with whitespace in it. And, you won't overflow your command line buffer.
The -print0 will use the NULL as a file separator instead of a newline and the -d $'\0' will use NULL as the separator while reading.
find . -name "*.txt"|while read fname; do
echo "$fname"
done
Note: this method and the (second) method shown by bmargulies are safe to use with white space in the file/folder names.
In order to also have the - somewhat exotic - case of newlines in the file/folder names covered, you will have to resort to the -exec predicate of find like this:
find . -name '*.txt' -exec echo "{}" \;
The {} is the placeholder for the found item and the \; is used to terminate the -exec predicate.
And for the sake of completeness let me add another variant - you gotta love the *nix ways for their versatility:
find . -name '*.txt' -print0|xargs -0 -n 1 echo
This would separate the printed items with a \0 character that isn't allowed in any of the file systems in file or folder names, to my knowledge, and therefore should cover all bases. xargs picks them up one by one then ...
Filenames can include spaces and even control characters. Spaces are (default) delimiters for shell expansion in bash and as a result of that x=$(find . -name "*.txt") from the question is not recommended at all. If find gets a filename with spaces e.g. "the file.txt" you will get 2 separated strings for processing, if you process x in a loop. You can improve this by changing delimiter (bash IFS Variable) e.g. to \r\n, but filenames can include control characters - so this is not a (completely) safe method.
From my point of view, there are 2 recommended (and safe) patterns for processing files:
1. Use for loop & filename expansion:
for file in ./*.txt; do
[[ ! -e $file ]] && continue # continue, if file does not exist
# single filename is in $file
echo "$file"
# your code here
done
2. Use find-read-while & process substitution
while IFS= read -r -d '' file; do
# single filename is in $file
echo "$file"
# your code here
done < <(find . -name "*.txt" -print0)
Remarks
on Pattern 1:
bash returns the search pattern ("*.txt") if no matching file is found - so the extra line "continue, if file does not exist" is needed. see Bash Manual, Filename Expansion
shell option nullglob can be used to avoid this extra line.
"If the failglob shell option is set, and no matches are found, an error message is printed and the command is not executed." (from Bash Manual above)
shell option globstar: "If set, the pattern ‘**’ used in a filename expansion context will match all files and zero or more directories and subdirectories. If the pattern is followed by a ‘/’, only directories and subdirectories match." see Bash Manual, Shopt Builtin
other options for filename expansion: extglob, nocaseglob, dotglob & shell variable GLOBIGNORE
on Pattern 2:
filenames can contain blanks, tabs, spaces, newlines, ... to process filenames in a safe way, find with -print0 is used: filename is printed with all control characters & terminated with NUL. see also Gnu Findutils Manpage, Unsafe File Name Handling, safe File Name Handling, unusual characters in filenames. See David A. Wheeler below for detailed discussion of this topic.
There are some possible patterns to process find results in a while loop. Others (kevin, David W.) have shown how to do this using pipes:
files_found=1
find . -name "*.txt" -print0 |
while IFS= read -r -d '' file; do
# single filename in $file
echo "$file"
files_found=0 # not working example
# your code here
done
[[ $files_found -eq 0 ]] && echo "files found" || echo "no files found"
When you try this piece of code, you will see, that it does not work: files_found is always "true" & the code will always echo "no files found". Reason is: each command of a pipeline is executed in a separate subshell, so the changed variable inside the loop (separate subshell) does not change the variable in the main shell script. This is why I recommend using process substitution as the "better", more useful, more general pattern.See I set variables in a loop that's in a pipeline. Why do they disappear... (from Greg's Bash FAQ) for a detailed discussion on this topic.
Additional References & Sources:
Gnu Bash Manual, Pattern Matching
Filenames and Pathnames in Shell: How to do it Correctly, David A. Wheeler
Why you don't read lines with "for", Greg's Wiki
Why you shouldn't parse the output of ls(1), Greg's Wiki
Gnu Bash Manual, Process Substitution
(Updated to include #Socowi's execellent speed improvement)
With any $SHELL that supports it (dash/zsh/bash...):
find . -name "*.txt" -exec $SHELL -c '
for i in "$#" ; do
echo "$i"
done
' {} +
Done.
Original answer (shorter, but slower):
find . -name "*.txt" -exec $SHELL -c '
echo "$0"
' {} \;
If you can assume the file names don't contain newlines, you can read the output of find into a Bash array using the following command:
readarray -t x < <(find . -name '*.txt')
Note:
-t causes readarray to strip newlines.
It won't work if readarray is in a pipe, hence the process substitution.
readarray is available since Bash 4.
Bash 4.4 and up also supports the -d parameter for specifying the delimiter. Using the null character, instead of newline, to delimit the file names works also in the rare case that the file names contain newlines:
readarray -d '' x < <(find . -name '*.txt' -print0)
readarray can also be invoked as mapfile with the same options.
Reference: https://mywiki.wooledge.org/BashFAQ/005#Loading_lines_from_a_file_or_stream
# Doesn't handle whitespace
for x in `find . -name "*.txt" -print`; do
process_one $x
done
or
# Handles whitespace and newlines
find . -name "*.txt" -print0 | xargs -0 -n 1 process_one
I like to use find which is first assigned to variable and IFS switched to new line as follow:
FilesFound=$(find . -name "*.txt")
IFSbkp="$IFS"
IFS=$'\n'
counter=1;
for file in $FilesFound; do
echo "${counter}: ${file}"
let counter++;
done
IFS="$IFSbkp"
As commented by #Konrad Rudolph this will not work with "new lines" in file name. I still think it is handy as it covers most of the cases when you need to loop over command output.
As already posted on the top answer by Kevin, the best solution is to use a for loop with bash glob, but as bash glob is not recursive by default, this can be fixed by a bash recursive function:
#!/bin/bash
set -x
set -eu -o pipefail
all_files=();
function get_all_the_files()
{
directory="$1";
for item in "$directory"/* "$directory"/.[^.]*;
do
if [[ -d "$item" ]];
then
get_all_the_files "$item";
else
all_files+=("$item");
fi;
done;
}
get_all_the_files "/tmp";
for file_path in "${all_files[#]}"
do
printf 'My file is "%s"\n' "$file_path";
done;
Related questions:
Bash loop through directory including hidden file
Recursively list files from a given directory in Bash
ls command: how can I get a recursive full-path listing, one line per file?
List files recursively in Linux CLI with path relative to the current directory
Recursively List all directories and files
bash script, create array of all files in a directory
How can I creates array that contains the names of all the files in a folder?
How can I creates array that contains the names of all the files in a folder?
How to get the list of files in a directory in a shell script?
based on other answers and comment of #phk, using fd #3:
(which still allows to use stdin inside the loop)
while IFS= read -r f <&3; do
echo "$f"
done 3< <(find . -iname "*filename*")
You can put the filenames returned by find into an array like this:
array=()
while IFS= read -r -d ''; do
array+=("$REPLY")
done < <(find . -name '*.txt' -print0)
Now you can just loop through the array to access individual items and do whatever you want with them.
Note: It's white space safe.
I think using this piece of code (piping the command after while done):
while read fname; do
echo "$fname"
done <<< "$(find . -name "*.txt")"
is better than this answer because while loop is executed in a subshell according to here, if you use this answer and variable changes cannot be seen after while loop if you want to modify variables inside the loop.
You can store your find output in array if you wish to use the output later as:
array=($(find . -name "*.txt"))
Now to print the each element in new line, you can either use for loop iterating to all the elements of array, or you can use printf statement.
for i in ${array[#]};do echo $i; done
or
printf '%s\n' "${array[#]}"
You can also use:
for file in "`find . -name "*.txt"`"; do echo "$file"; done
This will print each filename in newline
To only print the find output in list form, you can use either of the following:
find . -name "*.txt" -print 2>/dev/null
or
find . -name "*.txt" -print | grep -v 'Permission denied'
This will remove error messages and only give the filename as output in new line.
If you wish to do something with the filenames, storing it in array is good, else there is no need to consume that space and you can directly print the output from find.
function loop_through(){
length_="$(find . -name '*.txt' | wc -l)"
length_="${length_#"${length_%%[![:space:]]*}"}"
length_="${length_%"${length_##*[![:space:]]}"}"
for i in {1..$length_}
do
x=$(find . -name '*.txt' | sort | head -$i | tail -1)
echo $x
done
}
To grab the length of the list of files for loop, I used the first command "wc -l".
That command is set to a variable.
Then, I need to remove the trailing white spaces from the variable so the for loop can read it.
find <path> -xdev -type f -name *.txt -exec ls -l {} \;
This will list the files and give details about attributes.
Another alternative is to not use bash, but call Python to do the heavy lifting. I recurred to this because bash solutions as my other answer were too slow.
With this solution, we build a bash array of files from inline Python script:
#!/bin/bash
set -eu -o pipefail
dsep=":" # directory_separator
base_directory=/tmp
all_files=()
all_files_string="$(python3 -c '#!/usr/bin/env python3
import os
import sys
dsep="'"$dsep"'"
base_directory="'"$base_directory"'"
def log(*args, **kwargs):
print(*args, file=sys.stderr, **kwargs)
def check_invalid_characther(file_path):
for thing in ("\\", "\n"):
if thing in file_path:
raise RuntimeError(f"It is not allowed {thing} on \"{file_path}\"!")
def absolute_path_to_relative(base_directory, file_path):
relative_path = os.path.commonprefix( [ base_directory, file_path ] )
relative_path = os.path.normpath( file_path.replace( relative_path, "" ) )
# if you use Windows Python, it accepts / instead of \\
# if you have \ on your files names, rename them or comment this
relative_path = relative_path.replace("\\", "/")
if relative_path.startswith( "/" ):
relative_path = relative_path[1:]
return relative_path
for directory, directories, files in os.walk(base_directory):
for file in files:
local_file_path = os.path.join(directory, file)
local_file_name = absolute_path_to_relative(base_directory, local_file_path)
log(f"local_file_name {local_file_name}.")
check_invalid_characther(local_file_name)
print(f"{base_directory}{dsep}{local_file_name}")
' | dos2unix)";
if [[ -n "$all_files_string" ]];
then
readarray -t temp <<< "$all_files_string";
all_files+=("${temp[#]}");
fi;
for item in "${all_files[#]}";
do
OLD_IFS="$IFS"; IFS="$dsep";
read -r base_directory local_file_name <<< "$item"; IFS="$OLD_IFS";
printf 'item "%s", base_directory "%s", local_file_name "%s".\n' \
"$item" \
"$base_directory" \
"$local_file_name";
done;
Related:
os.walk without hidden folders
How to do a recursive sub-folder search and return files in a list?
How to split a string into an array in Bash?
How about if you use grep instead of find?
ls | grep .txt$ > out.txt
Now you can read this file and the filenames are in the form of a list.
Is there a way to get find to execute a function I define in the shell?
For example:
dosomething () {
echo "Doing something with $1"
}
find . -exec dosomething {} \;
The result of that is:
find: dosomething: No such file or directory
Is there a way to get find's -exec to see dosomething?
Since only the shell knows how to run shell functions, you have to run a shell to run a function. You also need to mark your function for export with export -f, otherwise the subshell won't inherit them:
export -f dosomething
find . -exec bash -c 'dosomething "$0"' {} \;
find . | while read file; do dosomething "$file"; done
Jac's answer is great, but it has a couple of pitfalls that are easily overcome:
find . -print0 | while IFS= read -r -d '' file; do dosomething "$file"; done
This uses null as a delimiter instead of a linefeed, so filenames with line feeds will work. It also uses the -r flag which disables backslash escaping, and without it backslashes in filenames won't work. It also clears IFS so that potential trailing white spaces in names are not discarded.
Add quotes in {} as shown below:
export -f dosomething
find . -exec bash -c 'dosomething "{}"' \;
This corrects any error due to special characters returned by find,
for example files with parentheses in their name.
Processing results in bulk
For increased efficiency, many people use xargs to process results in bulk, but it is very dangerous. Because of that there was an alternate method introduced into find that executes results in bulk.
Note though that this method might come with some caveats like for example a requirement in POSIX-find to have {} at the end of the command.
export -f dosomething
find . -exec bash -c 'for f; do dosomething "$f"; done' _ {} +
find will pass many results as arguments to a single call of bash and the for-loop iterates through those arguments, executing the function dosomething on each one of those.
The above solution starts arguments at $1, which is why there is a _ (which represents $0).
Processing results one by one
In the same way, I think that the accepted top answer should be corrected to be
export -f dosomething
find . -exec bash -c 'dosomething "$1"' _ {} \;
This is not only more sane, because arguments should always start at $1, but also using $0 could lead to unexpected behavior if the filename returned by find has special meaning to the shell.
Have the script call itself, passing each item found as an argument:
#!/bin/bash
if [ ! $1 == "" ] ; then
echo "doing something with $1"
exit 0
fi
find . -exec $0 {} \;
exit 0
When you run the script by itself, it finds what you are looking for and calls itself passing each find result as the argument. When the script is run with an argument, it executes the commands on the argument and then exits.
Just a warning regaring the accepted answer that is using a shell,
despite it well answer the question, it might not be the most efficient way to exec some code on find results:
Here is a benchmark under bash of all kind of solutions,
including a simple for loop case:
(1465 directories, on a standard hard drive, armv7l GNU/Linux synology_armada38x_ds218j)
dosomething() { echo $1; }
export -f dosomething
time find . -type d -exec bash -c 'dosomething "$0"' {} \;
real 0m16.102s
time while read -d '' filename; do dosomething "${filename}" </dev/null; done < <(find . -type d -print0)
real 0m0.364s
time find . -type d | while read file; do dosomething "$file"; done
real 0m0.340s
time for dir in $(find . -type d); do dosomething $dir; done
real 0m0.337s
"find | while" and "for loop" seems best and similar in speed.
For those of you looking for a Bash function that will execute a given command on all files in current directory, I have compiled one from the above answers:
toall(){
find . -type f | while read file; do "$1" "$file"; done
}
Note that it breaks with file names containing spaces (see below).
As an example, take this function:
world(){
sed -i 's_hello_world_g' "$1"
}
Say I wanted to change all instances of "hello" to "world" in all files in the current directory. I would do:
toall world
To be safe with any symbols in filenames, use:
toall(){
find . -type f -print0 | while IFS= read -r -d '' file; do "$1" "$file"; done
}
(but you need a find that handles -print0 e.g., GNU find).
It is not possible to executable a function that way.
To overcome this you can place your function in a shell script and call that from find
# dosomething.sh
dosomething () {
echo "doing something with $1"
}
dosomething $1
Now use it in find as:
find . -exec dosomething.sh {} \;
To provide additions and clarifications to some of the other answers, if you are using the bulk option for exec or execdir (-exec command {} +), and want to retrieve all the positional arguments, you need to consider the handling of $0 with bash -c.
More concretely, consider the command below, which uses bash -c as suggested above, and simply echoes out file paths ending with '.wav' from each directory it finds:
find "$1" -name '*.wav' -execdir bash -c 'echo "$#"' _ {} +
The Bash manual says:
If the -c option is present, then commands are read from the first non-option argument command_string. If there are arguments after the command_string, they are assigned to positional parameters, starting with $0.
Here, 'echo "$#"' is the command string, and _ {} are the arguments after the command string. Note that $# is a special positional parameter in Bash that expands to all the positional parameters starting from 1. Also note that with the -c option, the first argument is assigned to positional parameter $0.
This means that if you try to access all of the positional parameters with $#, you will only get parameters starting from $1 and up. That is the reason why Dominik's answer has the _, which is a dummy argument to fill parameter $0, so all of the arguments we want are available later if we use $# parameter expansion for instance, or the for loop as in that answer.
Of course, similar to the accepted answer, bash -c 'shell_function "$0" "$#"' would also work by explicitly passing $0, but again, you would have to keep in mind that $# won't work as expected.
Put the function in a separate file and get find to execute that.
Shell functions are internal to the shell they're defined in; find will never be able to see them.
I find the easiest way is as follows, repeating two commands in a single do:
func_one () {
echo "The first thing with $1"
}
func_two () {
echo "The second thing with $1"
}
find . -type f | while read file; do func_one $file; func_two $file; done
Not directly, no. Find is executing in a separate process, not in your shell.
Create a shell script that does the same job as your function and find can -exec that.
I would avoid using -exec altogether. Use xargs:
find . -name <script/command you're searching for> | xargs bash -c