I've been fiddling with some sed, most found through searching around the web, the 3 lines below work for me in that order one after another. I don't need the 3 backup files, if it can overwrite the original that could be great to.
Hope someone can help
Thanks
Matt
sed -i.bkp 's/^[ \t]*//'
sed -i.bkk '/^\s*$/d'
sed -i.bbb '/^[0-9]/ d'
What does this do for you? You may need to insert an 's' character after each semi-colon.
sed -i.bkp 's/^[ \t]*//;/^\s*$/d;/^[0-9]/d'
What does this do for you? You may need to insert an 's' character after each semi-colon.
sed -i.bkp 's/^[ \t]*//;/^\s*$/d;/^[0-9]/d'
That seams to work for me,
I can't remember but I think it first removes empty lines, then removes tabs/spaces at the start of lines then removes lines that begin with a number.
Thanks again
Matt
To overwrite the original, with no backup, use -i ""
To append commands, use -e
Try:
sed -i "" -e 's/^[ \t]*//' -e '/^\s*$/d' -e '/^[0-9]/ d'
Many Thanks that helps too!!
Just realised something else, I use this as a list for renaming, sometimes there are some illegal characters, namely there are some "/" can that be replaced with " - " can that be done for all illegal characters?
Many Thanks
Matt
Although the semi-colon can be used to string multiple sed commands together, it reduces readability, and the use of the -e gives your code breathing space, and added comprehension. You should use -e.
There are two ways to approach changing the '/' character to '-' using sed. You can escape the '/', or you can use different separators:
Using -e, you would simply append either syntax to your existing sed train.
The above is good if you know that you have just one character that needs transformed. Sed also supports changing a collection of characters to a single character, so by example the following will look for '/' and '*' (Sed reserved, must be escaped) and transform their occurrences to '-' for every processed line.
-e s'/[\/\*]/-/gp'
But as far as I know, Sed will not perform multiple replacements without multiple sequences of s/find/replace/g. The tr(1) utility on the otherhand can do many:many character mappings. Send the output of your existing Sed command sequence (train) into tr to perform multiple replacements:
sed train | tr '/*' '-x'
which would replace every occurrence of '/' with '-', and every occurrence of '*' with 'x' on each processed line. You are not limited to two characters.
-e s'/[\/\*]/-/gp'
I've opted for this one but in my script it works but duplicates the line.
I'm encapsulating it into applescript so looks like this
set thescript to "sed -i \"\" -e 's/^[ ]*//' -e '/^\\s*$/d' -e '/^[0-9]/ d' -e '/Introduction/d' -e '/Conclusion/d' -e s'/[\\/\\*]/-/gp' "
Drop the trailing p then. Don't you just love those AppleScript escapes to distract readability.
Hi VikingOSX,
Characters within a bracket expression do not need to be escaped but you have to be careful in which order you place special characters within the expression. Overall it is much safer to use POSIX character classes, ie
sed 's/[[:punct:]]/-/g'
I'm just glad thats the one little thing I have learnt took me ages to figure out how to make it work!
Mark,
I am fully aware of the POSIX class syntax, and should have used it, were it not for old habits that continue to haunt my fingers. For some reason, that I now cannot duplicate, sed blew up on me when I used unescaped characters in that bracket previously. That is why they were posted escaped. Had to have been some self-imposed typo.
Merge sed coding?
