escaping floats with backslashes works but doesn't get saved

[porres]

I found out you can escape a float with backslashes and you can create a symbol like that

the problem is that the backslash disappears from the message box when the patch is reopened

[umlaeute]

i think that \5 is not a valid way to express a symbol 5. please don't get in the habit of using it. (languages like Python or C would take \5 to mean  (aka the ENQUIRY character, or \u0005).

that said, I agree that not storing the slash in the patch is a bug (like the similar reports #1166, #1083). while superficially different, i think that all these bugs would be resolved if somebody would sit down and fix the escaping logic for good.

[danomatika]

I agree. I think we need simple, understandable escaping which works within Pd and via pdsend and pdreceive.

[Spacechild1]

languages like Python or C would take \5 to mean � (aka the ENQUIRY character, or \u0005).

That's a good point. However, Pd's escape sequences work quite differently than in those languages, because the set of special characters is not the same. For example, Pd doesn't need to escape ', ? or ". Nevertheless, I agree that repurposing \<number> could be confusing to people coming from those languages.

What about using an unused escape sequence, like \s, to force the following characters to be interpreted as a symbol, e.g. foo \s123?

Generally, there's a problem with saving new escape sequences, because it means that your patch might not work as intended when opening it on older Pd versions...

---

As a side note: Besides Pd's "special" characters (;, ,, $), we can only escape whitespace. I think \n (newline), \r (carriage return) and \t (tab) could be easily added to the list, as people coming from C like languages would naturally expect this to work.

In a distant future, we might even support \uhhh and \Uhhh for typing Unicode symbols :-)

[danomatika]

\s is a good idea.

[danomatika]

I'd recommend closing this issue and opening a new one specifically about improving escape mechanisms.

[umlaeute]

i don't know about closing: the original issue described here is that a string \5 (whether that's valid or not) doesn't get saved as such. i'm guilty as charged for derailing this into a "\5 is not a way to turn floats into symbols" discussion (on which i think we all agree by now).

so what should we do with a string \5?

[Spacechild1]

so what should we do with a string \5?

Different programming languages/runtimes handle invalid escape sequences differently. In the case of foo \s, Python3 would happily print foo \s, while the C/C++ standard libraries (that I've tested) would drop the backslash and print foo s, but they would also give a compiler warning.

Personally, I would opt for dropping the backslash in invalid escape sequences. Additionally, we could also print an error message, like "unsupported escape sequence ... in symbol ...`.

[umlaeute]

my question was more specifically geared towards \<d> with <d> being an integer.

but anyhow, my ~libc~ (EDIT:) C-compiler gives me:

string in the code	hex representation in the runtime	notes
\s	73 00	warning: unknown escape sequence: '\s'
\5	05 00
\25	15 00	025 == 0x15 == 21
\22.2	12 2E 32 00	022 == 0x12 == 18
\270	B8 00	0270 == 0xb8 == 184

i probably should have known that \<num> (specifically r"\[0-9]{1,3}") is used for the octal representation of a character.

[danomatika]

I like this. I think it makes sense to adopt something like C's conventions for this, when possible.

UPDATE: We are already doing this, to some degree, with the number representation, ie. %g.

UPDATE 2: I think we are doing this already, but it would be good to hide aspects of the implementation from this, ie. TCL-specific escaping should be hidden, where possible. Best to keep GUI implementation agnostic as much as we can IMO.

[porres]

where is this going? can I close this?