Re: [LUG] Sed experts wanted, perhaps?

[Philip Hudson <phil.hudson@xxxxxxxxx>]

On 19 Apr, 2013, at 9:36 pm, Grant Phillips-Sewell wrote:

> I'm trying to write a little script that can read a file and invoke
> twidge to tweet its contents to Twitter.
>
> The # symbol, being a special character (I think) is the root the my
> problems... and being Twitter, the # symbol is going to be used a fair
> bit.
>
> If I invoke twidge from the command line as this:
> $ twidge update "#generic #twitter #update"
> then all is well. The # symbol(s) are wrapped up in "" nicely.
>
> I haven't got to the reading-from-a-file bit yet, but here's what I  
> have:
>
> #!/bin/sh
> # Send Twitter updates through Twidge
> command="twidge update"
> status="$@"
> $command $status
>
> As it is, if I have no # symbols, it will work fine... however pop a #
> in there and it fails. It also fails with # symbols if I wrap the last
> line as:
> $command "$status"
>
> As it is, if I escape the # symbols, then it works fine... so I crack
> out sed...:
> status=`echo "$@" | sed -e "s/#/\#/g"` ... fails
> status=`echo "$@" | sed -e "s/\#/\\#/g"` ... fails (escaping the #
> symbol in sed)
> status=`echo "$@" | sed -e "s/\#/\\\#/g"` ... fails (escaping the #
> symbol and the \ symbol in the replacement string)
>
> Now, regexes of any nature have never been my strong point. I was
> successful with
> status=`echo "$@" | sed -e "s/~/#/g"` ... (replacing the ~ with the  
> # symbol)
> but I don't really want my text file(s) to all have ~ instead of #
> symbols if I can get away with it.
>
> Anybody got any ideas?

We've seen several good ones that should be getting you closer to a  
solution: check your quoting and escaping, try piping to stdin, use $*  
not $@, use variables not literals, replace # with ~ for intermediate  
steps...

Couple more, neither of them as useful as the foregoing, I fear:

1. (Optimization) Wherever you see 'sed s/x/y/g' where x and y are  
both single characters, think 'tr' instead.

2. Several of the foregoing tips are variations on the theme of  
indirection. Remember, in *n*x everything is a file. Rather than shell  
variables in RAM, try putting your data into temp files on disk at  
each step -- the shell is probably doing that behind the scenes if the  
data are non-trivial anyway. This gives you out-of-band diagnostic  
capabilities, which are very useful. Once you've got things working,  
optimize the files away again if you must.

> tr '#' '~' < raw_messages.txt > scrubbed_messages.txt
> head -1 scrubbed-messages.txt > scrubbed_message_1.txt
> twidge update < scrubbed_message_1.txt

Or something like that.

--
Phil Hudson                  http://hudson-it.no-ip.biz
@UWascalWabbit                 PGP/GnuPG ID: 0x887DCA63


--
The Mailing List for the Devon & Cornwall LUG
http://mailman.dclug.org.uk/listinfo/list
FAQ: http://www.dcglug.org.uk/listfaq