Python toolbox library updated

Those of you who attended the WROCC talk will be aware that the toolbox library for python has undergone a bit of an overhaul.

The reults of this are now available, and can be installed with pip in the usual way. You might need to tell pip you want to upgrade riscos-toolbox if you’ve been using 0.2.0.

A word of warning, things build for 0.2 will need a bit of fettling to use 1.0.1.

The Hyper example is on github (https://github.com/c-jo/riscos-toolbox) which should give you a good idea how to do stuff, given the lack of docs and howtos. Which are on the list of things to do!

Hi Chris
Thanks for doing this. Are you going to update PIP on PackMan. It is currently showing:- Version 20.0.2-4

Meanwhile I used PIP to upgrade itself but am a bit puzzled by the warning.

*python3 -m pip install pip -U
Collecting pip
  Downloading pip-23.1.2-py3-none-any.whl (2.1 MB)

Installing collected packages: pip

Successfully installed pip-23-1-2

WARNING: You are using pip version 23-1-2; however, version 23.1.2 is available.

You should consider upgrading via the

‘SCSI::SSD….!Python3.Bin.python38 -m pip install —upgrade pip’

command.

Replicating the above I decided to verify the upgrade had happened.

*python3 -m pip --version
Traceback (most recent call last):
....
  File "ADFS::Titan4.$.Progm.Progm2.Python.!Python3.python38.lib.importlib.resources", line 61, in _normalize_path
    raise ValueError('{!r} must be only a file name'.format(path))
ValueError: 'cacert.pem' must be only a file name

I remembered this.

A bit of digging found instances of cacert.pem in !PythonSite.Python38.Site-Packages.pip._vendor.certifi.core/py which are not present in the previous version of that file in pip 20.0.2 .

A speculative search and replace of all 7 .pem‘s with /pem’s helped.

**python3 -m pip --version
pip 23.1.2 from ... !PythonSite.Python38.Site-Packages.pip (python 3.8)
*

Perhaps PackMan versions are better for RISC OS??

A bit later Reverted to pip 20.0.2 from PackMan.

Excellent timing, just grabbed it for the project I’m working on, can’t wait to give it a shot.

All occurrences of ‘.’ used as an extension separator should be replaced with

os.path.extsep

I’m not sure if upstream would accept the patch, but it’s worth at least trying!

As to pip. It was a bit of an effort to get the version on PackMan working, so I suspect any installed from the net versions won’t work on risc os. As mentioned at the talk, I am going to try to port 3.12, which will, no doubt, be a challenge. I did get 3.10 to work, but I don’t think I ever tried 3.11.

The reults of this are now available, and can be installed with pip in the usual way. You might need to tell pip you want to upgrade if you’ve been using 0.2.0.

Just in case anyone else makes my mistake!
I took this to mean tell pip to upgrade itself. PIP upgrades ok but then stops working.
It means, tell pip to upgrade the toolbox.

As did David above I reverted to pip 20.0.2 from PackMan and used it to install the toolbox.
I am now having a go at getting the old toolbox example to work with the new toolbox.

It means, tell pip to upgrade the toolbox.

Sorry, that wasn’t very clear was it. I’ve upated the post to make it clearer!

I am struggling creating a suitable block which I can pass to Wimp_SendMessage in order to send a user message to another application. I created a class like this:

class HelpReplyMessage(UserMessage)
    event_id = 0x503
    _fields_ = [
        ("help_message", ctypes.c_char*236)
    ]

Then I can create the message, ready to send, like this:

message = HelpReplyMessage()
message.your_ref = 0
message.code     = 0x503
swi.swi("Wimp_SendMessage", "iII", Wimp.UserMessage, ctypes.addressof(message), task)

But I cannot for the life of me work out how to actually set that help_message string. I’ve searched the internet for ctypes examples, but almost all those involving strings have c_char_p, and the string is, of course, embedded in the message block.

Thanks to the efforts of Laren, there are a load of extra gadgets supported.

Version 1.0.2 available from pip.

Excellent, thank you Chris! I’ve been a bit busy the past couple days but I’m already in the process of porting something over to give the new stuff a real world test. Thanks again for your work!

Is this going to be merged back into main?

Chris: Thanks for all the work you’re doing on Python for RISC OS. I’m currently learning Python on Mac OS and Windows for work so really looking forward to trying on RISC OS when I have a bit more knowledge.

Lauren: I don’t know what gadgets are yet but they sound significant so thank you, as well!

Bumping this up to the top again in the hope that someone has a solution to this

I have searched for c_char in the Toolbox library code in the hope of finding an example of setting the contents of a c_char * NNN block with the contents of a string, but I think all the examples in the library are of reading strings!

Matthew: After the chat the other day I didn’t want to just leave this unanswered, so…

First, I think you’re overwriting the parent class’s fields definition there, so you may want to append it something along these lines:

fields = UserMessage.fields + [ (“help_message”,ctypes.c_char*236) ]

That said, I tried for hours and I couldn’t get anywhere either :( No matter what I tried I always got “Message block is too big / not a multiple of 4” and weirdly I noticed when I put a character array field in it always showed as being 20 bytes more than it was expected to be. But even if I adjusted for that it was the same thing…

I ran out of time, but hopefully either Chris can come back and answer this or if I ever figure it out in the course of things I’ll let you know… I’ll probably have to revisit it later.

First, I think you’re overwriting the parent class’s fields definition there

It looks like it, doesn’t it! I have no idea how is works, but that bit functions fine: I copied it from examples elsewhere in the Toolbox library, such as here, for example

As for the problem, it looks like I was over-thinking it. I cannot remember precisely what I tried last time, but just setting

message.help_message = b"Press this button to cause untold catastrophe"

seems to work. I should have had another go before bumping the thread, as I hit on it straight away. I’m sorry you struggled at it like me first time round!

My next problem is how to read two strings which are passed in the same buffer, one immediately after the other separated by a single null byte. I have only myself to blame for the message format, as it’s one I devised for Nominatim to maximise the capacity of a limited buffer. Dead easy to read in C or BASIC, but I’ve yet to work it out for Python!

I first thought of struct.unpack but that only does fixed length objects, but you could use it to extract the part of the buffer with the two null terminated strings and call

struct.unpack(spec, buffer).spit(b"\x00")[0:1]

There are a couple of problems with the python toolbox test program !Hyper.

1. The Keyboard short cuts in the menu do not work. My guess is they were not implemented.

2. The “Previous” button in the status line does not work for some cards. Eg:
The Toolbox – Object Templates
Tools
Tools.!DDT
Tools.ResEd
Tools.ResTest
What these cards have in common is that they have buttons. All the cards without buttons return to the previous card when previous is clicked. The cards with buttons do not.

If you delete the three buttons on the DDT card the “previous” or “back_handler” button works.

    def back_handler(self, event, id_block, poll_block):
        card = self.history.pop() if len(self.history) else None
        self._display(card)

It looks as if the history stack is being corrupted by processing the buttons defined on the card.

Remember, you can raise issues on the Python toolbox library in the usual way by posting them on the project itself – https://github.com/c-jo/riscos-toolbox/issues.

After the nice demonstration of using the ToolBox with Phyton, thanks to Chris, I undertook to make a version in C++ only with the DDE. It was a challenge with Paolo :-)

This program had always interested me and I wanted to understand how it worked, it’s very informative.
I have to clean up the sources and they will be available.
I put a copy of the demo directory, there is a “bug” in the hcl Tool_2 file. I think this is intentional, because this is the list of people who participated in the project.
To access it, use the search on the keyword “whodunnit”, you will receive an error message…
To see the list of participants you must remove the “;” in the last card, line 421.
Thank you for testing.
download

Buttons and shortcuts seem to work!

The toolbox is very good tool, for anyone who wants to create applications and it is available in many languages!!!

Nice, Jean-Michel, I’d like to see the source when ready. I’m honestly a big proponent of the Toolbox, having a proper toolkit sure beats doing things the old-fashioned way, which is why I was eager to try it out with Python.

Now that said, I’m very foolish so rather than clean up the source or make it anything near ready, I’m going to post what I’ve been making with it and the more recent version of the Python toolbox library; it’s mostly there to exist as another example or demonstration, so nobody take it as best practice. Also, it’s fully functional but completely useless to everyone except perhaps myself:

https://github.com/laurenrad/RLook

I’ll be working on it more soon, but consider this an extremely soft launch. It’ll keep getting updated alongside the library as well.

Remember, you can raise issues on the Python toolbox library in the usual way by posting them on the project itself – https://github.com/c-jo/riscos-toolbox/issues.

Thanks for reminder – Issue now raised on github.

@ Lauren
I have finished cleaning up the source of the Hyper2 program (C++), not easy, I hope you will excuse my mistakes in English. Please read the readMe

https://jeanmichelb.riscos.fr/Download/Hyper2Sce.zip

I’m honestly a big proponent of the Toolbox,

Me too, as all the libraries, I see that Pyhton is not lacking…

@ Jean-Michel Oh, there is an example file included within the RLook app bundle, by the way. I meant to pull it out of the app folder, but forgot.

Thanks Lauren,
This is a good demonstration of Rlook.
Interesting the content of these files and with your application you can listen to the samples, nice.