Skip to content

Latest commit

 

History

History
187 lines (111 loc) · 5.31 KB

shell_linux.md

File metadata and controls

187 lines (111 loc) · 5.31 KB
Raw

Misc

find which shell is in use echo $shell

For sourcing: .cshrc

For sourcing: .profile, .bash_profile or .bashrc

mkdir -p some/dir/sub/next/more create parents as needed

sed

sed -i 's/foo/bar/g' *

sed always does the first match per line, then moves to next line !!! -i is for --in-place which is changing the actual file. --in-place[=SUFFIX] keeps the original file with SUFFIX in the end of the filename, edits a new file (without SUFFIX)

Testing: sed 's/C1/SGEO_C1/g' <file.las feeds the file.las content to the sed command and outputs to std-out; then do the -i on the actual file

sed -i 's/C1/SGEO_C1/g' *.las

sed -i 's_ \{2,\}\t\t_"&_' *.txt

finds two or more spaces {2,}, then two tabs \t\t replaces the match & with a doublequote and the match "& Can also be included in a shell-script (Git Bash on Windows works nice)

#!/bin/bash
# changing strings in many las files in current directory
sed -i 's/C1/SGEO_C1/g' *.las

When in need to use a slash /, then use backslash to escape.

had to replace 48_ with GB48/ this seemed to work (Git Bash, Windows): sed -i 's/48_/GB48\//' *.las GB-wells: 2-step

Add blank where a number after minus sed -i 's/-[0-9]/-&/' *.txt then all the ones that should have minus and space have --, new sed: sed -i 's/--/- /' *.txt

Add to beginning of line sed -i 's/^/"/' test.txt

Find first tab, replace with doublequote and tab sed -i 's_\t_"\t_' file.txt

grep

find a string in files

grep SOR *.las look for SOR in all .las files, prints also the line where found

grep -l SOR *.las prints only the filename where found

grep -l SOR *.las | xargs.exe cp -t ./selected_curves/ to copy the found files to a different folder

file properties

Time

  • mtime modification time => ls -l shows that
  • ctime change of a file's metadata such as ownership, location, permissions
  • atime access time => try ls -lu

also find with -atime might help: To find files whose modification time is 2 or more days ago: find . -type f -mtime 2

ls

Oldest in the bottom: ls -lt

find

for files and folders/directories

find looks recursively from current pwd

  • or the given path find <pathtosearch> <options like name...>
  • can also be several paths in a space-delimited list find <pathtosearch> <anotherpath> <options like name...>

find -type f is for finding only files

find -type d is for finding only directories

find -name "*.txt" looks for patterns in name (glob), case-sensitive ==> without the quotes (not sure difference single vs. double) the shell would attempt to interpret *.txt

-iname would be case-insensitive

-path is like name but on the entire path. * will match both / and leading dots in filename

-ipath then case-insensitive

Each expression is true or false for the current file. Many those expressions are chained as -and per default. $ find -name "*.js" -type f would do the and for the name and type file.

-or would allow for several tests: find -name "*.js" -or -name "*.css" -type f

-not to negate expression

paranthesis group expressions, but require escaping in bash.

-exec to do something with the results

2>/dev/null to direct stderr to nowhere, thus not shown in terminal any more (e.g. access denied messages)

-ls is an option to show results (see man for sizes 1K usually)

nicer maybe: find ./BH21E_HA_40* -name "*_1445*" | xargs ls -lh then with human-readable sizes (more examples online for tricky cases)

Examples for find:

Find a specific filename in only some of the present directories find ./BH2* -name somefile.ext

regex

Regular expressions

https://github.com/google/re2/blob/main/doc/syntax.txt

rsync

-az archive and compress is good

-n for dry-run

--exclude attic/ to omit some parts

--bwlimit=300 limit the bandwidth used

trailing slash (/) on the source directory:

  • If you do not use a trailing slash, the source directory is copied to the destination directory, and then the contents of the directory.
  • When you do use the trailing slash, rsync only copies the content of the source without creating an additional directory level.

du

Disk usage

du -h to get human-readable size (like Megabyte, GB when appropriate), will traverse recursively into subdirectories and have one line per file

du -hs just output one summarized line with the total if the argument was a folder

du -h -d1 will summarize for the folders at folder-depth = 1

du -hs BH21E_HA_30* seems to work:

6.0T	BH21E_HA_300
90G	BH21E_HA_301
1.6G	BH21E_HA_302

ncdu

--confirm-quit prevents from unintentional close when hitting Esc too often

ncdu -e -o outfile.ncdu /some/path => can become quite large

=> open that saved file again: -f ncdu -f outfile.ncdu

shows top right corner that a file is open

cli

Environment variables: printenv display all

vs. Shell variables: Environment variables visible to all processes, not only shell

VAR="hello" # shell variable created export VAR # variable now part of the environment

aliases

alias newname='command'

ncdu, not to quit with q only: alias ncdu='ncdu --confirm-quit'

Users

id username shows numerical ids for user and groups

id -nu <number> shows the user to the numerical id

blocl of code => Ctrl-e in the browser editor

uniq for uniqueness...