Character escape codes

williamclocksin · December 15, 2024, 1:01pm

Why does Scala not have the \s escape code to mean space? It seems to have all the other escape codes.

JD557 · December 15, 2024, 1:32pm

I never saw \s and, from some quick tests, it doesn’t seem to be that common:

C: No
Javascript: No
Python: No
Java: Yes
Kotlin: No
Clojure: No

Other than Java, I couldn’t find any other language that supports that, including the other JVM competitors.

I also have to admit that I don’t understand the use case… Isn’t " " shorter and easier to read?
But maybe I’m missing something.

som-snytt · December 15, 2024, 4:55pm

It was a late addition in Java for text blocks.

I’m pretty sure there is a ticket to add it in Scala.

Oh right, now we have two issue trackers to check.

github.com/scala/scala3

support escaping space in string literals using `\s`

opened 06:04PM - 05 Apr 24 UTC

unkarjedy

itype:enhancement compat:java

Java multiline blocks supports `\s` escape sequence. https://docs.oracle.com/…en/java/javase/21/text-blocks/index.html > The \s escape sequence simple translates to space (\040, ASCII character 32, white space.) Since escape sequences don't get translated until after incident space stripping, \s can act as fence to prevent the stripping of trailing white space. Using \s at the end of each line in the following example, guarantees that each line is exactly six characters long. > >```java >String colors = """ > red \s > green\s > blue \s > """; >``` Scala doesn't support `\s` escape sequence. Scala 2 compiler crashes <img width="1069" alt="image" src="https://github.com/scala/scala3/assets/3989292/eae6548f-63ed-416b-91b3-2dcfa815bd8a"> Scala 3 compiler says > invalid escape '\s' not one of [\b, \t, \n, \f, \r, \\, \", \', \uxxxx] at index 4 in "aaa \s bbb". Use \\ for literal \. Given that `\s` is not reserved, the compiler could support by simply replacing with `\s` At first glance, this should be trivial to fix. This wouldn't work in non-interpolated multiline string literals and in raw string literals because they don't support escape sequences. Still, it could still be a nice addition. <img width="150" alt="image" src="https://github.com/scala/scala3/assets/3989292/e54898bb-44a5-4461-82eb-7775ba4ccb39">

JD557 · December 15, 2024, 5:09pm

Oh, I see: This is useful in Java because of the multiline string trimming, which differs from Scala.

I do see how that could be also useful in Scala, though. That makes more sense.

williamclocksin · December 15, 2024, 6:02pm

I like “\s” because it is instantly clear that a space is needed. Using “ “ always makes me wonder what precisely is between the quotes.

som-snytt · December 15, 2024, 9:53pm

As I just saw, someone suggested a test with regular offsets

"some code         // end of code"
"more code         // end"

where

"some code          \s"

is more descriptive.

The “terminal space” indicates exactly where the text block ends. A “strip margin” interpolator could also leverage that terminator.

sageserpent-open · December 16, 2024, 6:06am

Is this intended for all string literals, or just multi-line ones?

I’m thinking of folks whose unquoted regexes that use \s subtly change in meaning once this is rolled out.

som-snytt · December 16, 2024, 6:16am

No doubt there would be unsubtle warnings at every stage of migration.

The purpose of warnings is to make everyone unhappy with a change in behavior, whether or not they endorse it.

sageserpent-open · December 16, 2024, 7:00am

Hmm.

I’m overdramatising now, but when somebody’s library with their Acme Regex gets deployed in production, then some bright spark decides to upgrade the Scala distribution to pick up the latest bug fixes and there is no corresponding new version of said library to go with it, then won’t the only warning seen be the sudden volume of support requests as to why Form B30 doesn’t work anymore?

EDIT: Having stated that, why would any sane and responsible person deploy to production without running an assiduous set of automated tests first? No tests? Then they deserve all they get.

You have my blessing!

EDIT #2 The penny just dropped, the string interpolation occurs at compile time, so one would have to rebuild Acme Regex in order for this to be of consequence. At which point you get a warning. Doh!