Sed: what is the difference between -i and > file?

Question

I know that > file is not sed specific, and that it writes stdout to the file, but how does this compare to sed's -i option? I read the man page, and I do not understand what in-place means. Are either of these choices definitively better? Which is faster? Takes less ram? If you have any additional insights, please share.

Answer 1

It all depends what you want to do - but pipes are generally very fast and just uses IO buffers.

-i is a non-standard option to edit files in place. sed otherwise just transforms the data on standard in to standard out

On my machine I can't notice any speed difference on a 1mb file.

$ cat my_file.txt 
This is my file
$ sed -i "" s/my/your/g my_file.txt
$ cat my_file.txt 
This is your file
$

Answer 2

"In-place" means the operation uses space (like diskspace or memory) already taken, no additional resource is needed. In this context it's about sticking to the provided file.

The main advantage of sed -i -e '…' file is you can read file and have the output in a file still named file. It looks like the same file, but behind the scenes it's not. By default, sed reads the given file, processes it outputting into a temporary file, then moves the temporary file to the original name, replacing the original file.

So it's not strictly "in-place" because another file is needed and at some point you have two files; but the result is almost as if it was in-place. Writing to a separate file has its benefits:

The old version of the file is atomically replaced by the new version, so at every point in time the file name points to a valid, complete version of the file.

On the other hand sed -e '…' file >file doesn't work. Worse, it makes file lose its content. The behavior is explained here: Bending a pipeline back into its origin. If you try to achieve with > what -i does then you will most likely lose data.