In the 1970s, sed, a Stream EDitor, was created.
sed reads files line by line, is mainly used for search and replace, and
doesn’t edit the input file by default (instead printing the information to the
screen).
The options for sed change with each operating system, so I’ll stick the the
options for GNU (“GNU’s Not UNIX”, aka the precursor to Linux and what Linux
systems use).
Generic sed commands follow the syntax of:
$ sed 'script' input_file
So, if you wanted to search for every instance of the word “hello” and change
it to “world” in a file named input.txt, creating a new file with those
changes called output.txt, then the command would look like
$ sed 's/hello/world/g' input.txt > output.txt
If you wanted these changes to be reflected in the original file by overwriting
it, then you would instead use the -i flag, which edits in-place.
$ sed -i 's/hello/world/g' input.txt
In these examples, the s/ stands for search and the g stands for global;
together they are the equivalent of “find and replace all.”
Not including the g will make turn an example.txt file like
__hello__ my dearest friend __hello__
my deepest __hello__ unto you
can you reply to __hello__ with __hello__
I only say __hello__ to myself
into this
__world__ my dearest friend __hello__
my deepest __world__ unto you
can you reply to __world__ with __hello__
I only say __world__ to myself
because, once again, sed makes changes line by line.
Similarly, sed is not recursive, so you can globally replace a word with a
phrase containing the world multiple times without causing an infinite loop.
Now, single quotes are not always critical to the use of sed, but they won’t
hurt anything either.
Basically, they’re necessary for meta-characters (aka anything you shouldn’t
include in a file name).
Thus, since nobody understands sed, and the people writing sed
documentation suggest using single quotes every time, you should just learn
sed using single quotes every time.
You can also delete lines with sed.
To edit a file by removing the first line, then you would use
$ sed -i '1d' filename
In the example, 1d essentially stands for first line deletion.
Using a\ (“append”) can add lines to a file.
$ sed '/hello/ a\ Add this line after every line with hello' example.txt
hello my dearest friend hello
Add this line after every line with hello
my deepest hello unto you
Add this line after every line with hello
can you reply to hello with hello
Add this line after every line with hello
I only say hello to myself
Add this line after every line with hello
You can “insert” lines with i\
$ sed '/hello/ i\ Look at me go' example.txt
Look at me go
hello my dearest friend hello
Look at me go
my deepest hello unto you
Look at me go
can you reply to hello with hello
Look at me go
I only say hello to myself
or “change” lines with c\
$ sed '/my/ c\ well this was dumb' example.txt
well this was dumb
well this was dumb
can you reply to hello with hello
well this was dumb
These commands can easily be created into a script for editing files. Say you wanted to delete the CRYST line from all the PDB files in a folder. Hmmm, it’s almost like VMD hates this line from Avogadro-generated files. To do this, you could write a script like:
#!/bin/bash
sed -i '/CRYST/ c\' *.pdb
If you have symbolic links (see ln, you can have sed
follow the symbolic links. To do that, use something like:
$ sed -i --follow-symlinks 's/^/\t/' *.txt
In that example, a tab (\t) is inserted at the beginning of each line (^)
of every .txt file, including following through to the referenced files of
symbolic links.
Finally, old versions of sed only allowed the first line of a script to be a
comment, but now comments can be used anywhere in the script, though to be safe
they should be entered on their own line of the script.