Re: [Seed7-users] Q&A

[Thomas M. <tho...@gm...>]

On 2012-08-02 06:10 Kobi Lurie wrote:
> On 08/01/2012 04:51 PM, Thomas Mertes wrote:
>> In the meantime the existing documentation of the primitive actions at
>>
>>    http://seed7.sourceforge.net/manual/actions.htm
>>
>> can be used to see how a foreign function call can be done.
> ok, I'll look it up.
> I think I'll start by learning the language with simple exercises, 
> algorithms or data structures.

When you have questions, just ask.

>> then, later, probably contribute as I go along. (but let's not promise 
> anything ;-)

Okay.

>> 2) gui - is there a gui library?
>> Seed7 has a graphics library, but currently there is no gui library.
>> Except for the Gtk-server connection library provided by Leonardo
>> Cecchi. I have a plan to have an interface description and to use
>> this description for a web based interface and for stand alone
>> programs. But until I get massive support this will take very long,
>> to be realized.
> The graphics library looks really cool, from first impressions, it looks 
> suitable for lightweight games and small customized applications.
> However for commercial development, I assume most users like to see a 
> familiar UI, while the developer would like to not be held back by 
> unimplemented widgets, which is where the mature gui libraries come in.

Agree.
My attemts to write a gui library are described below. When you need
a gui library now I suggest you interface to an existing one.

>>> It seems portability in the language
>>> was a main concern, what do you feel about libraries such as fltk, or
>>> preferrably wxwidgets? (or is there a home-brewn one? do you wish for
>>> everything to be in one ecosystem?)
>> Interfacing with a library is okay. I don't know fltk and I have
>> only little knowledge about wxwidgets. Fitting a gui library fully
>> into the Seed7 ecosystem is probably not easy. Is it possible to base
>> a gui library on the Seed7 graphics library? I would like to avoid
>> event loops in the user program (I hate event loops). I also don't
>> like redraw events. It is just stupid when the program gets told:
>> I just deleted your window and you should redraw it.
>> What would programmers say, when this happens with files?
>> But this are just my wishes. When you do the work you are allowed
>> to ignore them. Feel free to add a cgi library to Seed7. I promise
>> to add it to the release, when it is not buggy and somehow useable.
> I don't know much about the web so probably cgi support will not come 
> from here.

Sorry I made a typo. I wrote cgi instead of gui. I wanted to say:
Feel free to add a gui library to Seed7.

> I feel similarly about events. Since they can "jump" up on you, you 
> don't know where they came from, they're not in the code flow, so it's 
> harder to reason and figure things out, especially for a larger program.
>
> However what is the alternative? since in a gui application all the 
> widgets are ready to accept input. they are like entities themselves. 
> How would you do that with one codeflow?

Maybe a system with input forms can be used. I am experimenting with
a gui library based on cgi forms. The form describtions look like
data structure descriptions. I would like a system where a user
interface definition is useable for stand alone applications and for
web (CGI) interfaces. But the realization will take a long time.
Don't wait for it.

> FWIW, If you're indeed going to implement a whole gui system, based on 
> the Graphics lib, I suggest looking at Rebol's gui dialect, which is the 
> nicest way I've seen for building GUIs. Moreover, maybe you can leverage 
> other's work for example cairo, agg, or clutter.
>
> Another promising option is QT's qml, which is an interpreted markup 
> language for ui. (similar to json in structure, css in definition, and 
> has some interaction, and conditional rules).
> I don't know, however, if you can really make large applications with 
> qml, and still be able to keep things under control.
> oh, there is also the way fantom do gui's which may be similar to what 
> you want to do on the web.
> they compile the application both to java for the desktop (or c#, 
> configurable) and to javascript for the web.

Thank you, for your suggestions.

>> To add 1 to each array element the expression
>>
>>    doMap(testArray, x, x+1)
>>    
>> can be used. The map function looks a little bit like a for-loop.
>> The variable x must be provided as parameter to doMap.
> hmm.. how does anExpression (inside the for loop) know to take aVariable 
> as its parameter?
> oh I see, they are just expanded to the external code x and the other x+1.
> that's quite smart.
>
> Is it possible to make map generic? instead of an integer -  it would 
> take a T, and return a T array?

Yes, it is possible. See:

  http://seed7.sourceforge.net/examples/generic.htm

I also added an example with a normal map function. See:

  http://seed7.sourceforge.net/examples/map.htm

>>> 4) auxiliary and fancy stuff -
>>> unit testing (and mocks),
>> Currently no special support, but I don't see problems.
>> The functionality of the Seed7 interpreter and compiler is checked
>> with several programs.
> I am sure the quality of your code is very high. It's my own feeble mind 
> I'm worried for :-)
> It helps me to check myself, since if the program gets complicated with 
> time, it's harder to find the source of a bug.

I know that you are talking about unit testing. I just want to
mention that compiled Seed7 programs can be debugged with a source
code debugger. Instructions in the generated C program refer to the
original source. This way a source code debugger like gdb can be
used.

> In recent years there is also 'tdd' (test driven design) where unit 
> testing has become a continuous and part of the coding process.
> mocks is about "faking" objects, usually by reflection, creating empty 
> functions and defaulted fields and storing the invocations, so for 
> example, you can test that your client code works, without connecting to 
> the server.
> since you know what the server is supposed to return to you, you "fake" 
> the server object, returning canned responses that you prepared, and 
> later it's possible to verify what was called, how many times, etc.
> Sometimes it helps, since you can get the program up and running fast, 
> and then fill all the holes with actual functionality. having a working 
> program is usually easier than waiting until all the code is finished 
> just to test it. Things can get complicated sometimes when a bug was 
> introduced but alot of code was written in one chunk, so which part is 
> good and which isn't?
> I think mocking mainly arose from object oriented and events, which can 
> get messy.
> An example library is moq. http://code.google.com/p/moq/

Thank you, for the hints.

>>> design by contract,
>> Currently no special support
>>> ADA's argument ranges,
>> Not sure what you mean with this. Seed7 supports positional
>> parameters and does not support named parameters.
> I saw that new version of Ada define ranges for the input parameters
> as seen here for example:
> http://en.wikipedia.org/wiki/Ada_(programming_language)#Data_types
> I think this is a good feature.
> probably just asserts in runtime, and is basically an invariant (like in 
> eiffel).
> that's good in order to know that your data didn't get corrupted somehow,
> and also (as you get the assert) you can know where that wrong data came 
> from.
> this is basically the advantage for dbc (design by contract) too.
> it's just a 'requires', that checks the input arguments. 'ensures' that 
> checks the result,
> and an invariant which isn't in a function but in the class, so it 
> checks the object's state wasn't compromised.

I see.

>>> mixins,
>> The functionality could probably be provided with Seed7 templates or
>> abstract data types.
>>> multiple inheritance,
>> Not supported
>>> events,
>> I already explained, that I don't think that event loops are a good
>> idea. But nobody hinders you to process events with Seed7.
>>> fine grained accessibility (not just public private but having syntax
>>> to limit to a function or class or with wildcards),
>> Fine grained accessability is currently not supported.
> Yes, this is somewhat experimental. didn't really expect it. more of a 
> feature request :-)
>>> actors and network programming ...
>> Not sure what you mean with this. Seed7 has support for networking
>> (FTP, HTTP) and CGI.
> I meant to ask about concurrency, threaded or actors-based, for example 
> in erlang there are lightweight processes, others use co-routines.
> about networking, I am assuming sockets.

Currently there is no threading / lightweight process support.
There is socket support. The other networking support is based on
socket support.

Regards,
Thomas Mertes