regexp search/replace not working as expected on individual authors in source field

[masp]

Hi, I've been using the "search and replace" (bulk edit) dialog with no problems for, but suddenly hit an unexpected behaviour. I'm using "authors" as my search field, and (my custom, multiple-value field) "editors" as destination. I want to copy a subset of the authors in each book as editors. e.g.:

authors: jim (ed) & bob & bill (ed) => editors: bill & jim

I tried the following regexps:

(.+?) \(ed\)
^(.+?) \(ed\)$

but in every case I'm seeing *all* authors copied, e.g.:

authors: jim (ed) & bob & bill (ed) => editors: bill & bob & jim

I can't figure it out. Any ideas? Thanks! Mike

[masp]

attached screenshot

[eschwartz]

(.+?) \(ed\)

Matches twice:

jim (ed)
& bob & bill (ed)

Try using a character set that does NOT include the ampersand:

([^&]+?) \(ed\)

[eschwartz]

I found http://www.regular-expressions.info to be a very helpful regex site, it taught me most of what I know.

I particularly appreciated their step-by-step guide through what the regex engine sees and does.

[masp]

Thanks! (I tried using [^&] but made no difference unfortunately.)

Actually my confusion stems not so much from the regexp syntax (I'm fairly familiar with that from my perl days) but because I was under the impression that when applying the regexp to a multi-value field (ie. an &-separated list of values, such as the "authors" field) the regexp would get applied to each individual value, rather to the whole string. In other words (pseudo-code), if field = "bob & jim (ed) & harry", then rather than this:

field.matches(regex)

it would be implemented more like this:

field.split("&").map { matches(regex) }

but that doesn't seem to be happening: did I misinterpret the docs?

[eschwartz]

On second thought, I am not sure what I thought you were thinking. Also, that will teach me to make assumptions.


Experimented a bit, and it looks like replacements are applied to the original string, which is then passed to the destination field. Also, you're right -- each value is treated separately.

So what you really want is to delete the non-editors, which I believe requires a negative lookaround.

[eschwartz]

Found it!

Search:

(^[^()]+?(?! \(ed\))$|\(ed\))

Replace: nothing. Don't replace with anything.


Strips out:

• any value in the field that doesn't have the string "(ed)" -- blue section
• the string "(ed)" -- red section

P.S. Sure is an ugly-looking regex.

[masp]

Aah! Thanks: I would never have figured that out!! Ugly but perfect in context

[chaley]

BTW: if you want to process a set of multiple items such as authors or tags as one string instead of individual items then use "template" as the source, enter a template that gives you the string, and enter the destination field. The attachment shows an example combining two authors into one. (I don't know why someone would want to do that, but it makes a good example.)

This technique is also useful for combining values from several fields into one.

[eschwartz]

Originally Posted by masp

Aah! Thanks: I would never have figured that out!! Ugly but perfect in context

My pleasure.

@chaley -- I took a stab at attempting some sort of template to solve the problem. Nothing jumped out at me, so I stuck with plain old field mode.

Even without split fields, you still have to modify the original string.