Streams - a beginning

[Bill Schwab]

Hello all,

I am open to suggestions, but I do not see a non-breaking way to get
stream behavior that is consistent with other dialects.  In particular,
to get #next to signal an error on exhaustion.  Squeak's #next: acts
like what others call #nextAvailable:.  IM(NS)HO, the others have it
right.  If a protocol calls for 41 characters to remain in a stream, I
want to be able to read 41, and would like to be warned if there are
fewer remaining.  If I do not care, #nextAvailable: allows me to
"authorize truncation."  Truncating silently can lead to the same kinds
of problems that (rightly) doomed "defensive programming."  Exceptions
offer the solution.

Attached is a beginning of an alternate stream heirarchy; it adds
classes vs. replacing them.  There are no file streams as yet.  One of
my design goals will be to get the user interaction out of the latter.  

My recommendation for working out of #next's Squeaky behavior is to add
#nextIfNone: that takes a block, and #nextOrNil.  It would then be a
relatively simple task to clean up many of the nil-checking sends of
#next.

Please take a look and let me know what you think.  Suggestions for
tests would be greatly appreciated.  If you can see a way to integrate
something like this into Squeak w/o pain, please let me know.

Bill



Wilhelm K. Schwab, Ph.D.
University of Florida
Department of Anesthesiology
PO Box 100254
Gainesville, FL 32610-0254

Email: bschwab@...
Tel: [phone]
FAX: [phone]



_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Bert Freudenberg]

On Nov 26, 2007, at 18:42 , Bill Schwab wrote:
> Please take a look and let me know what you think.

I did not take a look, but I am positive the UI list is the wrong  
place to discuss this.

- Bert -


_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Bill Schwab]

I respectfully disagree.  AFAICT, the I/O group is dormant, and the main
list has made its feelings clear.

However, there might be an argument for those interested to take this to
the I/O list.  Any takers?





Wilhelm K. Schwab, Ph.D.
University of Florida
Department of Anesthesiology
PO Box 100254
Gainesville, FL 32610-0254

Email: bschwab@...
Tel: [phone]
FAX: [phone]

>>> bert@... 11/26/07 12:47 PM >>>
On Nov 26, 2007, at 18:42 , Bill Schwab wrote:
> Please take a look and let me know what you think.

I did not take a look, but I am positive the UI list is the wrong  
place to discuss this.

- Bert -


_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[matthewf]

On Mon, Nov 26, 2007 at 01:00:31PM -0500, Bill Schwab wrote:
> I respectfully disagree.  AFAICT, the I/O group is dormant, and the main
> list has made its feelings clear.
>
> However, there might be an argument for those interested to take this to
> the I/O list.  Any takers?

I subscribed to the io list

the IO list may be appropriate, but the Flow list is probably
better. Note that there are currently two alternative stream
libraries in squeak (alternative meaning not in squeak.org)

1. Flow, which requires a plugin and supports many kinds of
   external streams, like USB streams and MIDI streams, in
   addition to collection streams and file streams. It is a
   major component of Spoon, and is not compatable with squeak's
   default stream library. See http://netjam.org/flow and
   http://netjam.org/spoon
2. Nile is a refactoring of the squeak stream library. It
   provides It does collection streams and file streams in two
   API's. The new API is much simpler than the squeak library,
   and is incompatable with it. The old API is compatable with
   squeak stream library. Nile uses traits to share code between
   the two API's. It does not provide any capabilities beyond
   the squeak libraries, as far as I can tell. See
   http://lists.squeakfoundation.org/pipermail/squeak-dev/2007-November/122481.html

--
Matthew Fulmer -- http://mtfulmer.wordpress.com/
Help improve Squeak Documentation: http://wiki.squeak.org/squeak/808
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Damien Cassou-3]

2007/11/27, Matthew Fulmer <tapplek@...>:

> On Mon, Nov 26, 2007 at 01:00:31PM -0500, Bill Schwab wrote:
> > I respectfully disagree.  AFAICT, the I/O group is dormant, and the main
> > list has made its feelings clear.
> >
> > However, there might be an argument for those interested to take this to
> > the I/O list.  Any takers?
>
> I subscribed to the io list
>
> the IO list may be appropriate, but the Flow list is probably
> better. Note that there are currently two alternative stream
> libraries in squeak (alternative meaning not in squeak.org)
>
> 1. Flow, which requires a plugin and supports many kinds of
>    external streams, like USB streams and MIDI streams, in
>    addition to collection streams and file streams. It is a
>    major component of Spoon, and is not compatable with squeak's
>    default stream library. See http://netjam.org/flow and
>    http://netjam.org/spoon
> 2. Nile is a refactoring of the squeak stream library. It
>    provides It does collection streams and file streams in two
>    API's. The new API is much simpler than the squeak library,
>    and is incompatable with it. The old API is compatable with
>    squeak stream library. Nile uses traits to share code between
>    the two API's. It does not provide any capabilities beyond
>    the squeak libraries, as far as I can tell. See
>    http://lists.squeakfoundation.org/pipermail/squeak-dev/2007-November/122481.html

Moreover, Nile is maintained by someone who gives the code to the
community and let it play with it and commit directly to the main
repository. Nile has a lot of unit tests too. If you want to play with
it, feel free. You can ask me any question.

Bye

--
Damien Cassou
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Bill Schwab]

Damien,

I have been aware of Flow for some time, but have reasons to avoid it,
at least for now.  When I tried it some time ago, Nile did not raise any
type of exception on reading off the end of a stream (and AFAIK, I was
using its streams vs. Squeak's).  I have since seen that is supposed to
work, so I will take another look.

Bill





Wilhelm K. Schwab, Ph.D.
University of Florida
Department of Anesthesiology
PO Box 100254
Gainesville, FL 32610-0254

Email: bschwab@...
Tel: [phone]
FAX: [phone]


>>> damien.cassou@... 11/27/2007 2:04 AM >>>
2007/11/27, Matthew Fulmer <tapplek@...>:
> On Mon, Nov 26, 2007 at 01:00:31PM -0500, Bill Schwab wrote:
> > I respectfully disagree.  AFAICT, the I/O group is dormant, and the
main
> > list has made its feelings clear.
> >
> > However, there might be an argument for those interested to take
this to

> > the I/O list.  Any takers?
>
> I subscribed to the io list
>
> the IO list may be appropriate, but the Flow list is probably
> better. Note that there are currently two alternative stream
> libraries in squeak (alternative meaning not in squeak.org)
>
> 1. Flow, which requires a plugin and supports many kinds of
>    external streams, like USB streams and MIDI streams, in
>    addition to collection streams and file streams. It is a
>    major component of Spoon, and is not compatable with squeak's
>    default stream library. See http://netjam.org/flow and
>    http://netjam.org/spoon 
> 2. Nile is a refactoring of the squeak stream library. It
>    provides It does collection streams and file streams in two
>    API's. The new API is much simpler than the squeak library,
>    and is incompatable with it. The old API is compatable with
>    squeak stream library. Nile uses traits to share code between
>    the two API's. It does not provide any capabilities beyond
>    the squeak libraries, as far as I can tell. See
>  

http://lists.squeakfoundation.org/pipermail/squeak-dev/2007-November/122481.html



Moreover, Nile is maintained by someone who gives the code to the
community and let it play with it and commit directly to the main
repository. Nile has a lot of unit tests too. If you want to play with
it, feel free. You can ask me any question.

Bye

--
Damien Cassou
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Damien Cassou-3]

2007/11/27, Bill Schwab <BSchwab@...>:
> I have been aware of Flow for some time, but have reasons to avoid it,
> at least for now.  When I tried it some time ago, Nile did not raise any
> type of exception on reading off the end of a stream (and AFAIK, I was
> using its streams vs. Squeak's).  I have since seen that is supposed to
> work, so I will take another look.

It still does not raise any exception because this is what ANSI says
IIRC. But I don't really care about ANSI anymore so if you want to
change the behavior of Nile please do and commit to the main
repository. Nile is designed to be easily changeable and extendable.

--
Damien Cassou
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Bill Schwab]

Damien,

I am far more interested in compatibility with VW and Dolphin (which
tends to follow VW in such areas) than with strict ANSI compatibility.
In the case of exhaustion, it sure seemed to me that ANSI did not
specify an error - I am convinced that Dolphin and VW have it right.

I will take a long look at Nile.  Thanks!

Bill




Wilhelm K. Schwab, Ph.D.
University of Florida
Department of Anesthesiology
PO Box 100254
Gainesville, FL 32610-0254

Email: bschwab@...
Tel: [phone]
FAX: [phone]


>>> damien.cassou@... 11/27/2007 9:38 AM >>>
2007/11/27, Bill Schwab <BSchwab@...>:
> I have been aware of Flow for some time, but have reasons to avoid
it,
> at least for now.  When I tried it some time ago, Nile did not raise
any
> type of exception on reading off the end of a stream (and AFAIK, I
was
> using its streams vs. Squeak's).  I have since seen that is supposed
to
> work, so I will take another look.

It still does not raise any exception because this is what ANSI says
IIRC. But I don't really care about ANSI anymore so if you want to
change the behavior of Nile please do and commit to the main
repository. Nile is designed to be easily changeable and extendable.

--
Damien Cassou
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Damien Cassou-3]

2007/11/27, Bill Schwab <BSchwab@...>:
> Damien,
>
> I am far more interested in compatibility with VW and Dolphin (which
> tends to follow VW in such areas) than with strict ANSI compatibility.
> In the case of exhaustion, it sure seemed to me that ANSI did not
> specify an error - I am convinced that Dolphin and VW have it right.
>
> I will take a long look at Nile.  Thanks!


Nile uses Traits and Traits are not available under other Smalltalk dialects.


--
Damien Cassou
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Bill Schwab]

Damien,

Understood, but the API can always be reproduced in terms of their
existing classes.  Is there a preferred forum for discussing Nile?

Bill




Wilhelm K. Schwab, Ph.D.
University of Florida
Department of Anesthesiology
PO Box 100254
Gainesville, FL 32610-0254

Email: bschwab@...
Tel: [phone]
FAX: [phone]

>>> damien.cassou@... 11/28/07 6:25 AM >>>
2007/11/27, Bill Schwab <BSchwab@...>:
> Damien,
>
> I am far more interested in compatibility with VW and Dolphin (which
> tends to follow VW in such areas) than with strict ANSI compatibility.
> In the case of exhaustion, it sure seemed to me that ANSI did not
> specify an error - I am convinced that Dolphin and VW have it right.
>
> I will take a long look at Nile.  Thanks!


Nile uses Traits and Traits are not available under other Smalltalk
dialects.


--
Damien Cassou
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Damien Cassou-3]

2007/11/28, Bill Schwab <BSchwab@...>:
> Damien,
>
> Understood, but the API can always be reproduced in terms of their
> existing classes.


Yes, and you can flatten the traits inside the classes.


> Is there a preferred forum for discussing Nile?

squeak-dev is ok I think. Since, I'm the only developper and there is
no user at all, we can also talk by mail :-).

--
Damien Cassou
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Bill Schwab]

Damien,

Fair enough.  Email will probably be easiest; Squeak-dev is busy enough
to swamp us.

Bill




Wilhelm K. Schwab, Ph.D.
University of Florida
Department of Anesthesiology
PO Box 100254
Gainesville, FL 32610-0254

Email: bschwab@...
Tel: [phone]
FAX: [phone]

>>> damien.cassou@... 11/28/07 9:49 AM >>>
2007/11/28, Bill Schwab <BSchwab@...>:
> Damien,
>
> Understood, but the API can always be reproduced in terms of their
> existing classes.


Yes, and you can flatten the traits inside the classes.


> Is there a preferred forum for discussing Nile?

squeak-dev is ok I think. Since, I'm the only developper and there is
no user at all, we can also talk by mail :-).

--
Damien Cassou
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[matthewf]

On Wed, Nov 28, 2007 at 10:17:03AM -0500, Bill Schwab wrote:
> Damien,
>
> Fair enough.  Email will probably be easiest; Squeak-dev is busy enough
> to swamp us.

I kindly request that you discuss this in public on squeak-dev.
It is pretty easy to just read mails on some threads and not
others, if you don't want to keep up with all squeak-dev
traffic.

--
Matthew Fulmer -- http://mtfulmer.wordpress.com/
Help improve Squeak Documentation: http://wiki.squeak.org/squeak/808
_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui

[Herbert König]

Hi,

MF> On Wed, Nov 28, 2007 at 10:17:03AM -0500, Bill Schwab wrote:
>>
>> Fair enough.  Email will probably be easiest; Squeak-dev is busy enough
>> to swamp us.

MF> I kindly request that you discuss this in public on squeak-dev.
MF> It is pretty easy to just read mails on some threads and not
MF> others, if you don't want to keep up with all squeak-dev
MF> traffic.

same with me and to Damien: publich discussion might even create users
:-))


Herbert                            mailto:herbertkoenig@...

_______________________________________________
UI mailing list
UI@...
http://lists.squeakfoundation.org/mailman/listinfo/ui