I new in pd and i'm trying to create an external that ouputs rich text. I don't found lot of thing about string in pd. String datatype can be really usefull if we want to create object like a mail reader, or xml reader.
In a more general way : what's about string manipulation in pd?

Thanks.

Jerome

Not good news for you, right now anyway. It's best to think of string processing in Pd as non-existant at present, otherwise you would have to admit that strings in Pd are a joke.

There is the concept of a symbol, but it's very limited. I do all my text processing in
Perl or Python, much better suited to the task.

Apparently Miller is working on new features to "improve" string handling.. Lists of symbols seem the obvious structure that fits in with the current model and could
remain compatible without introducing a new type "string", but alas symbols are
the black sheep of Pd which is essentially a tool for handling numeric data, so there
are many inconsistencies and problems with symbols.

If you want to build parsers and mail clients you will at very least need regular expressions and basic tools like strlen, chop, split, head/tail, rotate, reverse, alphabetical sort etc. I find that you can use the [shell] object to pass stuff out to
grep and sed, but it's not a pretty way to work and very inefficient. There's a few
goodies like [makefilename] and [tosymbol] that will save your ass when you
absolutely have no choice but to work with strings in Pd. But...
I hate to have to say it, and you're the first person I've had to say it to, but my sincere advice is that Pd is not the environment you're looking for. Not yet anyway.

There is a language written **specifically** for text and string processing and there
is no other language that comes close to it, not by a long way despite what badly informed advocates of other tools may tell you - that is the Practical Extraction and Reporting Language, otherwise known as Perl. Writing a mail server or XML parser
are all *solved* problems and you just have to drop in a module like XML::Parse, add
a few lines of code and the job is done.

If you are a Python programmer and can't bear to make the switch to Perl have a look at Twisted. This is a general "function junction", an object oriented playground for every imaginable network transaction. With Twisted you can build applicatins in no time, and they will scale up to enterprise size, play nicely with Zope/Plone to
easily become pluggable web-content objects.

arg....

In fact i'm using pda (gige.xdv.org/pda/) on a embedded device. Basic objects are enought for what I need, excepted string manipulation. Perl python, ruby... seems to be too slow for an embedded device.

When I read your post I understand that adding a new data type seems to be hard to do.

Thanks

There's nothing stopping you from writing entirely in low level form though. If you have absolutely no space availabe on your embedded device for anything but Pda then that might be your only option. It's just hard work yes. I would convert eveything characterwise into ints, there's [atoi] and [ord] objects somewhere that will let you build routines to convert between strings and lists of ints. Only problem is that they aren't ints, Pd only deals with floats. Don't even think about doing
any arithmetic on the data.In theory you could write your own alpha sort or whatever.

Google for "Puredata Text Ascii string" and you won't see much that satisfies,
there's http://www.ling.uni-potsdam.de/~moocow/projects/pd/ the externals by moocow. Maybe some more by now, but they are all hacks to get around the fact
that Pd has no "string" type at all.

But seriously, if text processing is the main function of your application I would choose a better tool. If it's just an incidental requirement then there's a hack you can use, but don't expect it to easy or that it will survive for long without getting broken by something new.

So, I decide to "use the source"... code
I tried tonight to add a new data type to the pd's source code. I choose it as basic char*. There are still some bug, and there are no memory management (must be done by the external itself), and no graphic components to enter or view the text. But it seems to be possible to add a new data type in pd (need to modify 6 or 7 pd's source code files and being patient during 4hours;) ). In the same way, integer can be add, but it's probably more complicated because integer array have to be define.

I feel that adding a new data type in pd is the best way to play with string (if it is in the official release of course). "ascii float" don't respect the data type security and can lead to "hard to understand" errors in the patch.

I'm trying to drive robot with an event driven and parallel system as pd. Sensor and actuator needs only numerical data, but speech to text, speech recognition, web interaction need text processing.

Great stuff!! Perhaps I could suggest you take this to the pd-dev list, they don't bite (much) and you might avoid any wheel reinvention as it;s been talked about a lot before. Some people might even be sitting on code that is still "unofficial".

My most basic desires for Pd strings would be a couple of proper ascii file objects that can work like IOstream, lock a stream/file, seek to a location etc. Look at the source for [qlist] and [textfile] to see the current offerings, they are both a bit perculiar. An alphanumeric comparison is also needed before you can move forwards with any sorting too. The big one I think is to get some kind of regex
internally in Pd, something like [grep] would be a nice object! I think once you have char any lines can just be taken as EOL terminated lists of chars. There is currently no way to create lists of lists in Pd either afaik, so you couldn't easily represent the entire file as a null terminated lol, but there are things called data-structures, which are not entirely unlike C structs, but once again not exactly what you think.

The problem you then have (political) is to convince the maintainers of other components, like the

operators, to accept and include support for your new
type. The two active developers I would discuss this with are Mathieu B and Miller himself.

Your robotics project sounds very interesting btw!

Hey Bilock,

I asked about on dev to get a bit more clued up here. Before you get too into modiying the source check out the pd string libs by Bryan Jurish (that's moocows site).

Strings are stored as byte lists with this method (which is 1/4 as efficient as it could be but you could also store UTF-8/16)

The disadvantages are that it's not easy to see or create strings this way without quite a lot of messing about converting things, but at least the representation is sound.

all best,

Andy

Thanks for your help. I'm still working on this pd's string, and i'm going to have a look to Bryan Jurish lib.

I'm trying also to use ruby script in pd, it's not easy but there's lot of things in gridflow for pd.

I'll bring this up for a new problem:

I'm working on reading out parameters from a vst.
I managed to split the space-separated attribute string using [unpack] but now I'm stuck.

Is there any possibility to extract the float value of such a string:
0.9980Amount

Ah, just discovered [stripfilename], I think that could work out.

What selector preceeds the data? In other words what does it say if you just print out the 0.9980Amount you have?

I could only get it out of unpack as a symbol.
But I just found the solution.
Thanks for your help though: I already learnt so much just from reading tons of posts of you.