ansaurus

Question

Answer 1

+3 A:

Use different delimiter @ for example

echo $HTML | sed -e s@ENTRY@$TABLE@

Anton 2010-05-17 14:25:48

That didn't work on my FreeBSD box however until I've put it in apostrophes `sed 's@ENTRY@$TABLE@'` or echoed a dollar sign `sed s@ENTRY@\$TABLE@`.

Yasir Arsanukaev 2010-05-17 14:40:17

It works for the sample above, but it failed when processing an entire html file.

StackedCrooked 2010-05-17 14:41:35

@StackedCrooked: How did it fail?

Anton 2010-05-17 14:45:26

With exactly the same error message: "unterminated `s' command".

StackedCrooked 2010-05-17 14:51:36

@StackedCrooked: Your variable contains character that is used as sed delimiter. Try using a delimiter char that is not used inside your file.

Anton 2010-05-17 15:15:48

FYI I updated my post with a code sample.

StackedCrooked 2010-05-18 08:02:14

@StackedCrooked: This time you have a problem with newlines in replacement. This fixed it:ENTRIES=`cat gui_template | sed -e s/FILENAME/test/ | sed -e s/TITLE/title/ | sed 's/$/\\\\n/' | tr -d '\n'`If you are not stuck to bash it maybe easier to use some other scripting language.BTW in case you don't know you can use #!/bin/bash -xto see what bash is really doing.

Anton 2010-05-18 08:41:20

Answer 2

+1 A:

Issuing these lines on FreeBSD console:

HTML="<html><body>ENTRY</body></html>"
TABLE="<table></table>"
echo $HTML | sed -e "s#ENTRY#$TABLE#"

Result in:

<html><body><table></table></body></html>

Yasir Arsanukaev 2010-05-17 14:25:51

Thanks that works fine.

StackedCrooked 2010-05-17 14:43:34

Oops, sorry I jumped to conclusions too fast. With the single quotes the variables aren't dereferenced anymore.

StackedCrooked 2010-05-17 14:53:53

@StackedCrooked: `echo $HTML | sed -e "s#ENTRY#$TABLE#"` works for me.

Yasir Arsanukaev 2010-05-17 14:58:45

It seems to work like this like this in a smaller sample, but not in the final script which processes a bigger html file. I'll do some more research..

StackedCrooked 2010-05-17 15:10:12

Answer 3

A:

You need to use a delimiter that can't appear in $TABLE, and if $TABLE is unpredictable enough this can be tricky. I'd suggest using a nonprinting character as a delimiter; it's easier to find one that's not going to show up in $TABLE and break everything. The only problem is they're harder to type in, so I'd suggest putting it in a variable and using that in the sed command:

DELIM=$'\377'
HTML="<html><body>ENTRY</body><html>"
TABLE="<table></table>"
echo "$HTML" | sed -e "s${DELIM}ENTRY${DELIM}$TABLE${DELIM}"

Note that the $'...' construct is a bash-only feature; if you need this to run under generic sh you'll have to do something messier, like DELIM="$(printf "\377")". Also, I chose \377 (that's FF in hex) because it's illegal in the UTF-8 encoding, so it should be safe if you're using UTF-8 for your HTML; if you're using something else, like Windows-1252, then \177 (the 'DEL' character) might be a safer choice.

Oh, yeah, and if you ever try to debug this with bash -x, be prepared for comedy.

Gordon Davisson 2010-05-17 23:49:49

I still can't get it to work, I updated my post with a code sample.

StackedCrooked 2010-05-18 08:03:21

ansaurus

tags:

views:

answers:

Using sed with html data

related questions