Any interest in raw strings? (original) (raw)

Stephen Colebourne scolebourne at joda.org
Wed Mar 25 03:41:59 PDT 2009

Personally, I would say that raw strings (prefixed by r, not multiline) were a good candidate for a Coin. The change would seem relatively small, and they compile to a standard String instance.

I wouldn't especially like to see RegExp strings, as it ties the compiler too closely to one specific library.

Stephen

2009/3/25 Reinier Zwitserloot <reinier at zwitserloot.com>:

Because raw strings are only used for regexps (it's so bad that people think the 'r' in python's raw string syntax stands for regexp and not raw), I suggested a while ago, on a different forum, to support regexp literals, which would even get the benefit of compile-time checking the regexp (there are a few things which are neccessarily a bad reg- exp, such as mismatched parentheses), and possibly the compile-time generation of the regexp literal to a regexp Thompson's tree, which takes relatively long to build, but once built, can parse any string through the regexp in O(n), where n is the length of the string.

Back then there were some admittedly rare use-cases for raw strings (including concatenating a regexp with a variable, which you couldn't do with regexp literals), but there was no consensus on whether raw strings, or regexp literals (or possibly both) were a good idea. I'd love me some raw (and multiline) strings, but this stuff is a bit too tricky to sort out now, a month or two before java7RC1.  --Reinier Zwitserloot

On Mar 25, 2009, at 00:58, Vilya Harvey wrote: Ah, I had a feeling I might have missed something. It looks like you've already covered everything I wanted to write up and more! I think multiline strings would be a useful addition too; it seems a shame that it won't be included in the Coin changes. Cheers, Vil.

2009/3/24 <rssh at gradsoft.com.ua> Hi all,

I don't think I've seen a proposal for raw strings - string literals which ignore escape sequences - yet. Would there be enough interest to write this up as a proposal? The idea is stolen shamelessly from Python, where you can write a raw string _as r"[\n\r\t]" and it's equivalent to the normal string "[\n\r_ \t]" (i.e. you don't need to double the backslash in a raw string, it's just treated like any other character). This is most useful when writing regular expressions; in fact, it may only be useful then. But for that one case it is extremely handy. Any opinions?

It was part of my enhanched string proposal  http://redmine.gradsoft.ua/wiki/java7stringliterals If creating new proposal only for unescaped strings (without multiline) can help (ask Darcy) -- than submit one. It woud be better than nothing. Vil.

More information about the coin-dev mailing list