Shell Programming and Scripting

I'm in the process of being forward-thinking and finally converting my site's db to UTF-8. I've already done the UTF-8 conversion (on a copy for testing) and now I want to go through and convert html entities to their actual characters.

I ran an entity decode on a mysqldump file but realized that entites for apostrophes and quotes in the file were converted to ' and " without the slashes, thereby screwing up mysql's field delimiting.

So I had the bright idea of searching for and replacing , say, each apostrophe html entity with a backslash and the entity. That way when I ran the html enity decode I'd end up with a properly escaped apostrophe.

I did

Code:

sed 's/[apos]/\[apos]/g' test.sql > withslashes.sql

and it didn't add the backslash.. (I'm using apos here because the board is interpreting the entities)

I then doubled the backslash to escape it, like this:

Code:

sed 's/[apos]/\\[apos]/g' test.sql > withslashes.sql

However that gave me this in the new file \aposapos

How do I insert a backslah with sed?

Thanks,

Do you mean something like:

Code:

echo "bla 'blabla\" bla" | sed "s/['\"]/\\\&/g"

If not, please provide input and output samples

What is [apos]? Are you looking for the literal string [apos] or somethign else?

For quoting quotes I am more used to

Code:

echo 'bla '\''blabla" bla' | sed 's/['\''"]/\\&/g'

that even works with a C-shell.
Rule: prefer the safer 'ticks' over the "quotes";
replace each ' within 'ticks' with '\''
so it becomes a \' escaped by the shell (outside the 'ticks').

I'll clarify my question. I found it difficult to post because my searches included html entities and the forum software was translating them into characters. The apos was just meant to represent the entity in the question.

Ok, so...

Search phrase: & #39; <--obviously without the space.
Replace phrase: \& #39; <-- backslash & #39; again obviously without the space.

Ahah. & has a special meaning to sed -- on the right side, it means 'the pattern that was matched', which is why you end up getting it twice. So you need to escape it, too, to have it taken literally.

In addition you can use the & in its meaning of "the pattern that was matched":

Code:

sed 's/&#39/\\&/g'

The first \ is to escape the second \