as some folks may remember, I have been working on making Python and its tool base more accessible to disabled programmers. I've finally come up with a really simple technique which should solve 80% of the problem. What I need to figure out is how to find a spot in the code where a symbol exists and potentially, it's rough type (class name, instance, etc.). This is really a much bigger question that I want to get into right now but I'm looking just to build a demo to back up a storyboard plus video.
When you are sitting on or in a name, you look to the left or look to the right what would you see that would tell you that you have gone past the end of that name. For example
a = b + c
if you are sitting on a, the boundaries are beginning of line and =, if you are sitting on b, the boundaries are = and +, if you are sitting on c, the boundaries are + and end of line. a call the region between those boundaries the symbol region.
if this example is clear to you, what you suggest for a method of finding a whole statement, or a whole symbol region? note, doesn't have to be perfect or complete solution, just good enough to let me do a moderately complex demo and seek funding accessibility world to build a complete environment.
I appreciate the help because I believe that once this is working, it'll make a significant difference in the ability for disabled programmers to write code again as well as be able to integrate within existing development team and their naming conventions.
On Sun, 29 Jul 2012 19:21:49 -0400, Eric S. Johansson wrote:
> When you are sitting on or in a name, you look to the left or look to
> the right what would you see that would tell you that you have gone past
> the end of that name. For example
Have you read the docs? It gives full details of the Python syntax.
> On Sun, 29 Jul 2012 19:21:49 -0400, Eric S. Johansson wrote:
>> When you are sitting on or in a name, you look to the left or look to
>> the right what would you see that would tell you that you have gone past
>> the end of that name. For example
> Have you read the docs? It gives full details of the Python syntax.
Yes I have. I was hoping for a different perspective because what I'm trying to do is middle out parsing. Top-down when the scanner focus moves from left to right and bottom up when the scanner focus moves from right to left.
sounds kind of odd when I describe it that way but both the cursor is on the middle of a name string and I need to look to either end of that name string before can do a conversion to a symbol string, I have to look at both ends in different ways. If you've read the documentation I've provided, would it be a better example to use for describing some of the issues. Here's a very rough draft of a storyboard
the first 13-14 slides are the working content for the storyboard. the rest is mostly "memory" of things I was thinking about so if it doesn't make sense or seems wrong, don't give me grief. :-)
> Here's a Python parser using the pyparsing library. It's a bit old
> (written for Python 2.4) but it shouldn't be hard to update it to new
> syntax:
> When you are sitting on or in a name, you look to the left or look to
> the right what would you see that would tell you that you have gone
> past the end of that name. For example
> a = b + c
> if you are sitting on a, the boundaries are beginning of line and =,
> if you are sitting on b, the boundaries are = and +, if you are
> sitting on c, the boundaries are + and end of line. a call the region
> between those boundaries the symbol region.
Check the lexical definitions. They essentially define, what
a "symbol region" is.
In essence, you have names, operators, literals whitespace and comments --
each with quite a simple definition.
> I would first tokenize the code, then divide it by statement keywords. > Finally, you just need to find expression/assignment statements in the > remaining sections. (Maybe there is a better way to do it.)
yeah the problem is also little more complicated than simple parsing of Python code. For example, one example (from the white paper)
*meat space blowback = Friends and family [well-meaning attempt]
*could that be parsed by the tools you mention? I suspect not but this is what I need to generate using speech recognition because it's easily spoken. A more complex example might be something like
new base = OS path-base name (old path)
or
if OS base exists (current path): new base name = OS path base name(current path)
What's particularly cute here is that using the translation technique I can actually describe the full object method path with a minimum of speaking overhead. Python is great. :-)
But the questions remain, will these tools are stuff like this?
> yeah the problem is also little more complicated than simple parsing > of Python code. For example, one example (from the white paper)
> *meat space blowback = Friends and family [well-meaning attempt]
> *could that be parsed by the tools you mention?
It is not valid Python code. Pygments is able to tokenize code that is not valid Python code. Because it is not parsing, it is just tokenizing. But if you put a bunch of random tokens into a file, then of course you will never be able to split that into statements.
Probably, you will need to process ident/dedent tokens, identify the "level" of the satement. And then you can tell what file, class, inner class, method you are staying in. Inside one "level" or code block, you could try to divide the code into statements.
Otherwise, I have no idea how a blind person could navigate in a Python source. In fact I have no idea how they use regular programs. So I'm affraid I cannot help too much with this. :-(
>> yeah the problem is also little more complicated than simple parsing of >> Python code. For example, one example (from the white paper)
>> *meat space blowback = Friends and family [well-meaning attempt]
>> *could that be parsed by the tools you mention?
> It is not valid Python code. Pygments is able to tokenize code that is not > valid Python code. Because it is not parsing, it is just tokenizing. But if > you put a bunch of random tokens into a file, then of course you will never be > able to split that into statements.
If you have been reading the papers, you would understand what I'm doing. I'm trying to take Python code with speech recognition friendly symbols and translate the symbols into a code friendly form. My conjecture is that you can change your perspective on the code and look for the edge that would normally be used to define start of a symbol, you should be able to define the name string. Another possibility is looking at the region which just contains letters numbers and spaces and outside and use that as your definition of a name string. It would probably help to verify that each word is found in a dictionary although that adds extra complexity if you are trying to increase the dictionary at the same time as the translation table.
I'm beginning to think for the first generation I should just use regular expressions looking forwards and backwards and try to enumerate the possible cases.
> Probably, you will need to process ident/dedent tokens, identify the "level" > of the satement. And then you can tell what file, class, inner class, method > you are staying in. Inside one "level" or code block, you could try to divide > the code into statements.
I was starting in that direction so that is good confirmation
> Otherwise, I have no idea how a blind person could navigate in a Python > source. In fact I have no idea how they use regular programs. So I'm affraid I > cannot help too much with this. :-(
I'm sorry, I am, and I'm trying to help, hand disabled programmers. There are more disability than blindness and after almost 20 years of encountering this shortsightedness, I do get a little cranky at times. :-)
On Mon, 30 Jul 2012 11:40:50 -0400, Eric S. Johansson wrote:
> If you have been reading the papers, you would understand what I'm
> doing.
That is the second time, at least, that you have made a comment like that.
Understand that most people are not going to follow links to find out whether or not they are interested in what you have to say. If you can't give a brief explanation of what you are doing in your email or news post, many people aren't going to read on. Perhaps they intend to but are too busy, or they have email access but web access is restricted, or they've already got 200 tabs open in their browser and don't want any more (I'm not exaggerating, I know people like that).
People use email because it is a "push" technology -- you don't have to go out and look for information, it gets pushed into your inbox. Clicking on links is a "pull" technology -- you have to make the explicit decision to click the link, open a browser, go out to the Internet and read who knows what. That requires a different frame of mind. Expect to lose some of your audience every time you require them to follow a link.
And *especially* so if that it a link to Google Docs, instead of an normal web page. Google Docs is, in my opinion, a nasty piece of rubbish that doesn't run on any of my browsers. As far as I'm concerned, I'd rather download a Word doc, because at least I can open that in OpenOffice or Abiword and read it. Something in Google Docs might as well be locked in a safe as far as I'm concerned.
> On Mon, 30 Jul 2012 11:40:50 -0400, Eric S. Johansson wrote:
>> If you have been reading the papers, you would understand what I'm
>> doing.
> That is the second time, at least, that you have made a comment like that.
Actually, it's probably more like the forth hundred time. :-) I apologize, I was wrong and I would back up and start over again if I could
> Understand that most people are not going to follow links to find out
> whether or not they are interested in what you have to say. If you can't
> give a brief explanation of what you are doing in your email or news
> post, many people aren't going to read on. Perhaps they intend to but are
> too busy, or they have email access but web access is restricted, or
> they've already got 200 tabs open in their browser and don't want any
> more (I'm not exaggerating, I know people like that).
accept criticism. I'm still working on an elevator pitch for this concept. I've been living with the technology and all its variations for about 10 years and it's not easy to explain to someone who is not disabled. People with working hands don't understand how isolating and, sometimes humiliating software can be. advocates like myself sometimes get a little tired of saying the same thing over and over and over again and people who are disabled just don't care. So you find yourself using shorthand because you going to be ignored anyway
> People use email because it is a "push" technology -- you don't have to
> go out and look for information, it gets pushed into your inbox. Clicking
> on links is a "pull" technology -- you have to make the explicit decision
> to click the link, open a browser, go out to the Internet and read who
> knows what. That requires a different frame of mind. Expect to lose some
> of your audience every time you require them to follow a link.
Okay, this implies the need to really work on more of an elevator/summary speech. Thank you for your input. I appreciate it
> And *especially* so if that it a link to Google Docs, instead of an
> normal web page. Google Docs is, in my opinion, a nasty piece of rubbish
> that doesn't run on any of my browsers. As far as I'm concerned, I'd
> rather download a Word doc, because at least I can open that in
> OpenOffice or Abiword and read it. Something in Google Docs might as well
> be locked in a safe as far as I'm concerned.
the ability for multiple people to work on the same document at the same time is really important. Can't do that with Word or Libre office. revision tracking
in traditional word processors are unpleasant to work with especially if your hands are broken.
It would please me greatly if you would be willing to try an experiment. live my life for a while. Sit in a chair and tell somebody what to type and where to move the mouse without moving your hands. keep your hands gripping the arms or the sides of the chair. The rule is you can't touch the keyboard you can't touch the mice, you can't point at the screen. I suspect you would have a hard time surviving half a day with these limitations. no embarrassment in that, most people wouldn't make it as far as a half a day. I've had to live with it since 1994. Not trying to brag, just pointing out the facts.
I'm going to try again from a different angle in a different thread. I will take your advice to heart and I would appreciate some feedback on how well I do satisfying the issues you have described
> the ability for multiple people to work on the same document at > the same time is really important. Can't do that with Word or > Libre office. revision tracking in traditional word processors > are unpleasant to work with especially if your hands are broken.
If you're developing, I might recommend using text-based storage and
actual revision-control software. Hosting HTML (or Restructured
Text, or plain-text, or LaTeX) documents on a shared repository such
as GitHub or Bitbucket provides nicely for accessible documentation
as well as much more powerful revision control.
> It would please me greatly if you would be willing to try an > experiment. live my life for a while. Sit in a chair and tell > somebody what to type and where to move the mouse without moving > your hands. keep your hands gripping the arms or the sides of
> the chair. The rule is you can't touch the keyboard you can't
> touch the mice, you can't point at the screen. I suspect you
> would have a hard time surviving half a day with these
> limitations. no embarrassment in that, most people wouldn't make
> it as far as a half a day.
I've tried a similar experiment and am curious on your input device.
Eye-tracking/dwell-clicking? A sip/puff joystick? Of the various
input methods I tried, I found that Dasher[1] was the most
intuitive, had a fairly high input rate and accuracy (both
initially, and in terms of correcting mistakes I'd made). It also
had the ability to generate dictionaries/vocabularies that made more
appropriate/weighted suggestions which might help in certain
contexts (e.g. pre-load a Python grammar allowing for choosing full
atoms in a given context).
<python.l...@tim.thechases.com> wrote:
> I've tried a similar experiment and am curious on your input device.
> Eye-tracking/dwell-clicking? A sip/puff joystick? Of the various
> input methods I tried, I found that Dasher[1] was the most
> intuitive, had a fairly high input rate and accuracy (both
> initially, and in terms of correcting mistakes I'd made). It also
> had the ability to generate dictionaries/vocabularies that made more
> appropriate/weighted suggestions which might help in certain
> contexts (e.g. pre-load a Python grammar allowing for choosing full
> atoms in a given context).
A microphone with voice recognition software is described in the storyboard.
> On 07/30/12 21:11, Eric S. Johansson wrote:
>> the ability for multiple people to work on the same document at
>> the same time is really important. Can't do that with Word or
>> Libre office. revision tracking in traditional word processors
>> are unpleasant to work with especially if your hands are broken.
> If you're developing, I might recommend using text-based storage and
> actual revision-control software. Hosting HTML (or Restructured
> Text, or plain-text, or LaTeX) documents on a shared repository such
> as GitHub or Bitbucket provides nicely for accessible documentation
> as well as much more powerful revision control.
But then you hit a second layer of "doesn't really work nice with speech recognition". Using a markup language can actually be more difficult than using a WYSIWYG editor. For example, with a soft word, I can do most of the basics using speech commands and I have what's called "Select-and-Say" editing capability in the buffer. Can't do that with any other editor. They are not integrated with NaturallySpeaking.
A few years ago I created a small scale framework for speech recognition users (akasha). You use the domain specific market notation to be able to construct Web applications. Instead of using classic HTML models, it used things that were more appropriate to speech driven environment. unlike with HTML, which you cannot write using speech recognition and a boatload of effort, akasha was 95% speakable using out-of-the-box speech recognition.
This also brings me to the concept of how the design for speech recognition use. Modifying an existing user interface or creating a new one either through a speakable data format and minimal application changes or by the application and grammar to a recognition engine and manipulating something that isn't speakable.
From experience, outside of the first model works well if you are looking for relatively easy and very high marks accessibility, the second is required if you are operating within a team and need to integrate with everybody else. Unfortunately, the second technique points out just how badly designed most software is and that led me to the concept of no-UI (not unlike no SQL) which is more controversial than I want to get into right now.
>> It would please me greatly if you would be willing to try an
>> experiment. live my life for a while. Sit in a chair and tell
>> somebody what to type and where to move the mouse without moving
>> your hands. keep your hands gripping the arms or the sides of
>> the chair. The rule is you can't touch the keyboard you can't
>> touch the mice, you can't point at the screen. I suspect you
>> would have a hard time surviving half a day with these
>> limitations. no embarrassment in that, most people wouldn't make
>> it as far as a half a day.
> I've tried a similar experiment and am curious on your input device.
> Eye-tracking/dwell-clicking? A sip/puff joystick? Of the various
> input methods I tried, I found that Dasher[1] was the most
> intuitive, had a fairly high input rate and accuracy (both
> initially, and in terms of correcting mistakes I'd made). It also
> had the ability to generate dictionaries/vocabularies that made more
> appropriate/weighted suggestions which might help in certain
> contexts (e.g. pre-load a Python grammar allowing for choosing full
> atoms in a given context).
Just ordinary speech recognition. NaturallySpeaking. Part of my hand problem is that I no longer have good fine motor control which sucks because I used to enjoy drawing with pencils and juggling. I've tried dasher and I don't have find enough motor control to make it work very well. Sometimes I play games with my girlfriend at fly or die and the user interfaces for the various game controllers is simple enough that my hands only get in the way some of the time. Or at least that's what say when she is beating me soundly at billiards. :-)
Some of the ideas you've mentioned have been thought in another contexts. The problem is that when it comes to working with code, you have two problems. Creation of code and editing code. Which do you do more? If you're like most of us, it is editing. That's why I made the toggle words feature something that would toggle both from a string name to a codename and vice versa.the future version of this and, this is where I'm going to need a lot of help from the Python community, would translate the statement in a buffer to a particular two-dimensional form and place it into a special window that NaturallySpeaking can operate on. The reason for covering it from the one-dimensional string of a statement to a two-dimensional form is to make it easier to disambiguate different features using speech. The idea isn't fully fleshed out because I want to get through this one first and actually be able to start writing code again.
<steve+comp.lang.pyt...@pearwood.info> wrote:
> Google Docs is, in my opinion, a nasty piece of rubbish
> that doesn't run on any of my browsers. As far as I'm concerned, I'd
> rather download a Word doc, because at least I can open that in
> OpenOffice or Abiword and read it. Something in Google Docs might as well
> be locked in a safe as far as I'm concerned.
I go the opposite way. Google Docs works fine in my web browser, but
if it's a Word doc, I need to hunt down something that can read it.
I've yet to find any browser that can't handle a GDocs "publish" page,
but I have plenty of computers that don't have any
{Open|Libre|Liber|Libra|whatever the next generation is}-Office
installed.
Best is to put the information into your email/post. Next best is to
have a link to the information. Definitely worst is to force people to
download your file and try to read it.
> On 7/30/2012 9:54 PM, Steven D'Aprano wrote:
> It would please me greatly if you would be willing to try an experiment. > live my life for a while. Sit in a chair and tell somebody what to type > and where to move the mouse without moving your hands. keep your hands > gripping the arms or the sides of the chair. The rule is you can't touch > the keyboard you can't touch the mice, you can't point at the screen. I > suspect you would have a hard time surviving half a day with these > limitations. no embarrassment in that, most people wouldn't make it as far > as a half a day.
Just using speech? Probably more people than you might think have had such experiences: anyone who's done software support over the telephone for a start! And in that scenario, they are effectively 'blind' too.
