Tuesday, June 04, 2019

Abolishing SyntaxError: invalid syntax ...

... and other cryptic messages.

Do you remember when you first started programming (possibly with Python) and encountered an error message that completely baffled you? For some reason, perhaps because you were required to complete a formal course or because you were naturally persistent, you didn't let such messages discourage you entirely and you persevered. And now, whenever you see such cryptic error messages, you can almost immediately decipher them and figure out what causes them and fix the problem.

Congratulations, you are part of an elite group! Even a large number of people who claim that they can program are almost certainly less capable than you are.

Given your good fortune, would you mind donating 5 to 10 minutes of your time to help countless beginners that are struggling in trying to understand Python error messages?  All you need to do is:

  1. Glance through of exceptions on this page and use your experience to find a case not covered. Note that this excludes SyntaxError cases, some of which are included here, but would require more of your time.
  2. Fire up your favourite Python REPL and write some simple code that generates an exception not already covered.  Perhaps, something like ValueError: could not convert string to float: 'a'
  3. Create a new issue with the error message as the title, including the code that generated the exception in the description of the issue together with a simple explanation (in a couple of sentences) of what the error message means.  Imagine that you are writing an explanation for the twelve year old child of your best friend who has expressed some interest in learning how to program.  This simple explanation is the most important part ... however, do not worry about getting it absolutely perfect as it will likely be improved upon based on feedback from future "real beginners".
  4. Go back to whatever you were doing before, knowing that the few minutes you have invested will cumulatively save many hours to future generation of programmers that encounter the exception you wrote about.
It should go without saying that contributions that require more time and effort that what is described above are also very welcome!  If you feel particularly ambitious, you can certainly improve the existing code that currently analyses cases of SyntaxError: invalid syntax, which currently handles only a few cases, and should be seen more as a prototype/proof-of-concept.

Future plans for friendly-traceback

Friendly-traceback is being written so that it could be easily incorporated into editors or IDEs that are designed for beginners. I intend to ensure that it can be easily added to Python's IDLE, as well as Mu and Thonny. I hasten to add that Thonny already includes an excellent tool (its "Assistant") which provides amazing feedback to beginners in some cases. Thonny's assistant uses a complementary approach to that of Friendly-traceback and it is quite likely that a future version of Friendly-traceback will include and expand upon the type of analysis performed by Thonny's assistant to help beginners. However, for the moment, the current development of Friendly-traceback is focused on breadth of coverage (i.e. increasing the number of exceptions included), providing a single most-likely explanation for each exception, rather than looking at multiple possible causes for a given exception as is done by Thonny's Assistant.

Friday, May 24, 2019

Avant-IDLE: an experiment

[Edit: this post has an embedded video, which is not visible for PlanetPython readers.]

This is a follow-up from a previous post where I showed just a screenshot made after one hacking session.  A week later, much has changed.

As noted at the end of the video, the code is not (yet) available in a public repository.  Among other things, I need to figure out what license I can use and if I can reuse all of the content from Python's version, or if I need to remove certain assets, etc.

Avant-IDLE makes use of two projects that I started recently and mentioned before on this blog:

Friendly-traceback (code, documentation) and AvantPy (code, documentation). AvantPy itself depends on Friendly-traceback.

Both Friendly-traceback and AvantPy welcome contributions.  Publishing this video about AvantPy is a way to bring more attention to these other two projects.

Going forward with Avant-IDLE

As for Avant-IDLE, if I can go forward with it:

  • I would like to translate its menus.
  • I would like to add syntax coloring for each individual dialect: this would likely mean some significant change from the way that syntax coloring is done in IDLE.
  • I would like to have friendly tracebacks make use of color, like iPython does.
  • I would like to add line numbers to the file editor.
  • etc.
I might have to remove the debugger, as it makes little sense to have someone entering code in their native language and suddenly have to deal with standard Python since that is the code that is truly executed.  I think that, by the time someone is ready to use a debugger, they are definitely ready to use Python itself.

In an ideal world, save for the "dialect converter" demonstrated in the video, I would like for the rest of Avant-IDLE to be in a single window, with files in tabs and the interpreter below, like Al Sweigart described:

However, that last idea might be something simply too ambitious. After all, Al Sweigart, who is far more competent and productive than I am, never managed to get that project off the ground. The more the code base from Avant-IDLE diverges from that of IDLE itself, the more difficult it will become to incorporate code changes made by Python's core developers to improve IDLE.

Making the code public

Even if I figure out all I need to do to make Avant-IDLE public, I am somewhat hesitant in doing so.

There is no doubt in my mind that I can continue working and improving both Friendly-traceback and AvantPy.  I truly believe that Friendly-traceback could be very helpful for Python beginners.  It could be integrated in other beginner-friendly editors, like Mu or Thonny - two fantastic projects.

I think that AvantPy could be useful for beginners as well ... but I do admit that it is a bit quirky.

When it comes to Avant-IDLE itself, as the saying goes, I don't want to bite more than I can chew, and find out that I created something which I cannot maintain. I do realize my limitations: I am just a hobbyist with no formal training in programming and who likes to do quirky experiments in his spare time. Publishing code on Github and/or making it available at Pypi automatically raises people's expectations, and demands on one's time.

Still, if this video can inspire you to create something useful for beginners, it will have been worthwhile.  Even more so if it inspires you to contribute to either Friendly-traceback or AvantPy, ;-)


If you have any constructive criticism, or suggestions to offer, please feel free to do so, either on this blog, or by email.

Friday, May 17, 2019

Idle hacking - part 1

  1. File test_py.pyfr.  The pyfr extension indicates that this is written in the "French Python" dialect. Since this is the main file we'll run, it will also automatically set the language to French. This is all done by AvantPy.
  2. "afficher" is the French version of "print"
  3. "import" is the normal Python keyword; we can mix and match keywords from a given dialect with the normal Python ones.  Here, we are importing a module whose base name is "upper".
  4. The constant "a" is defined.
  5. File "upper.pyupper"; the only file whose base name is "upper" - thus, the one that will be imported. The "pyupper" extension indicates that the dialect is an "upper case" version of Python - designed for testing.
  6. The code that will be run when the file is imported.
  7. Output from both files.
  8. "a" is defined.
  9. We made a syntax mistake. However, Friendly-traceback is there to help us understand what went wrong ... picking up the default language to use (French) from that first file that was run.

All this was run with a crudely hacked version of Idle. I'm still confused with the communications done with the RPCServer and the TPCServer, having never looked at similar code before.  So, I haven't been able to make it do all that I wanted in this first hacking session. Hopefully, there will be more to come ...

Wednesday, May 15, 2019

Friendlier tracebacks

When beginners run programs that generate Python tracebacks, they are almost always confused by what the information shown and have no clue as to what this all means. More experienced programmers can sometimes extract enough information directly from tracebacks to figure out what what wrong, but they will often have to resort to inserting a few print calls and running their program again to truly figure out what went wrong and how to fix it. (A minority of programmers might eschew using print calls and use a debugger instead.)

In order to make tracebacks more useful for them, some advanced programmers have designed tools to add more information so that simply looking at the enhanced traceback might be sufficient to diagnose properly the problem.  These tools include better_exchook, infi.traceback, rich-traceback, stackprinter, as well as the beautiful better-exceptions, and many others including Python's own cgitb module.  While there is no doubt that the additional information provided by these tools is useful for advanced programmers, it would likely only add to their confusion if it were used by beginners.

Friendly-traceback, a project I have been working on, and mentioned briefly in an addendum of my last post, aims at improving the situation for beginners.

A quick example

Have a look at the following Python traceback that might be generated by code written by a beginner.

Compare this with the following when using Friendly-traceback's own REPL

The only thing that is shown in exactly the same way is the line showing the exception name an a message. Everything else can be made available in other languages as shown below with a French translation.

Currently, Friendly-traceback can offer a more helpful message than a normal Python traceback in approximately 50 different cases, most of which are cases of either TypeError or SyntaxError. My goal is to eventually include explanations for all standard Python exceptions, and include as many sub-cases as possible.

As shown above, it is possible to translate the information in any language. Currently, only English and French are included; inclusion of other languages will require the help of volunteers.

If you write programs with your own custom exceptions, it is possible to write them so that they could be interpreted correctly by Friendly-traceback.  I have done so in another project, AvantPy, which I mentioned previously here. AvantPy includes 10 custom exceptions.

Both AvantPy and Friendly-traceback are available from Pypi. Contributors are most definitely welcome.

Friday, April 05, 2019

AvantPy needs you

Update: Since this blog post was published, and before Pycoder's weekly mentioned it bringing in quite a few visitors, I had decided to carve out the friendlier tracebacks into a separate project. If you are interested in this idea, you can have a look at the design document and feel free to comment.  Work on AvantPy itself has been paused until friendly-traceback's API is completely implemented.  The API itself is easy to do - the tedious work is adding in an explanation for every possible Python exception, and translating each of them ...

Imagine that you are either learning Python or helping someone that is learning Python. It is almost a certainty that you will see some Python tracebacks, often much more complicated to decipher than

Imagine if you could see something like the following instead:

Or, if you speak French:

Friendlier tracebacks, translated into various languages, is only one of many things that AvantPy can do to help beginners learn programming.

AvantPy is very much in its infancy. You can think of it as a proof-of-concept that can be built upon.

PSF's request for proposal

Yesterday, on the PSF's blog, some information about a request for proposal has been posted.
I have jotted down some thoughts about AvantPy in this context. If you wish to add your own thoughts on this, you can do so here. I do not wish to get any direct funding from the PSF for this project. However, some potential collaborators might need some financial support depending on their goals.

What is AvantPy

Here are some useful links if you want to find out more about AvantPy.

Contributors to the AvantPy project are very much welcome.