2015-08-20 10:19 GMT+02:00 stepharo <stepharo@free.fr>:
Le 20 août 2015 09:15, "stepharo" <stepharo@free.fr> a écrit :
For example why protocol get the same space than method. We could split
the protocol by half and use the space for
soemthing useful.
Protocol names can be long...
yes but 99% of the time you do not care.
I'm not sure of that. On classes with a lot of methods (Object, Morph, any class with over 20 methods), I believe one rely on protocols to avoid being overwhelmed by the sheer number, and that most of them are irrelevant to the task at hand (the method you're writing and reading, and the same class methods it is using). I wonder if the fact Nautilus shows us by default the 'all' method list is not inciting us, by pure laziness, to disregard protocols...
And methods are way more important. In fact protocol could just the tag (as in the catalog browser) after the method names.
Then you could use the message list GUI instead of the system browser. Thierry
Stef
I would like the coding area to expand a bit when a method is too long. I do it all the time by hand.
Yes, especially with Seaside renderOn:
Phil
Stef
Le 19/8/15 13:13, Thierry Goubier a écrit :
Hitting send too early :(
2015-08-19 11:59 GMT+02:00 Esteban Lorenzano <estebanlm@gmail.com>:
On 19 Aug 2015, at 11:50, Esteban Lorenzano <estebanlm@gmail.com>
wrote:
On 19 Aug 2015, at 11:44, Thierry Goubier <thierry.goubier@gmail.com>
wrote:
<Screen Shot 2015-08-19 at 11.10.41.png>
⦠but arriving to it is not so easy.
In conclusion: We are doing some right steps. It is not finished,
but we are not going to go back to older way :)
Hi Esteban,
My opinion is that you're still tied a lot to the Smalltalk 80's
way... Which is good: someone coming from 1980 would be able to use Nautilus ;)
My critics on that design: the tabs are nice and certainly help see
that a class has a class side. Overall look is more up to date. Tabs headers, scroll bars, etc... take far too much space: work area (the code area) is 39% of overall window size, and 68% counting in the context (package, class, protocol and method) (Numbers are worse on a small window, of course, but I guess some do work on small screens).
It will be nice to see simpler / cleaner Nautilus code coming along
:)
yes, of course you are right :) the GTools guys are working in a complete replacement, and Iâm sure it will be a lot better⦠but we will always need a backdoor⦠and I would like to have a good browser even as a backdoor.
I'm certainly looking forward for the GT-based system browser results, while probably I won't be using it; some of the design / implementation choices of the GT tools just don't fit the way I work, and I feel down by just looking at how clumsy and slow I become when using them.
(also, our philosophy is incremental: we improve what we have while
we wait for the break-thru improvements)
Agreed. I still miss a bit the OmniBrowser I started Pharo with... Especially the code regularity and architecture.
Esteban
ps: for me the âcode areaâ is not equivalent to the âwork areaâ: I
spend much more time understanding a problem than coding it, and for that a view of the method *in the context* is better)
That's why I added the 68% result: I do agree that the context matter; I still think that the overhead is too high.
for example: I try your alt browser time to time, because I find it
has some good ideas.
But since my workflow is usually: I dig a package, then I see the classes inside and try to figure out how they work together, then I read the comments, try to find examples, tests⦠then I finally go to the method level, and even that often with an eye into the class it belongs and even the package⦠so the AltBrowser is useless for me⦠it is really not confortable for me to use⦠and yes, it has a bigger code area, but that is not really relevant (for me). ⦠and the traditional browser adapts a lot better to that way of doing.
I think I more or less do the same, but with a slightly different approach, with for example a lot of scoping (to a class, to a package, implementors, senders, finder). I especially rely on the fact that I stay in the browser task model (but with multiple windows) all the time while doing that; I don't have to deal with disparate tools (Spotter, Finder, MessageList).
I also like the fact I can see both class and instance side methods at the same time...
Now⦠real question is: I work that way because fits better my
mind-model or my mind-model fits what the browser provides me? well, who knows⦠but back in my java days I also was using the âjava browsing perspectiveâ of Eclipse, who resembles a Smalltalk browser.
The good question we have here is that our search/exploration/understanding strategies are different, but based on the same building blocks... So there is some Moldable issue there as per the GT group, but not inside the tool as they are currently doing, instead the tool itself has to be moldable (or what is the tool).
Now, I'm so attuned to the Alt Browser tree view that I prefer it in any case... even to the point that I can do short and fast smalltalk edits directly on github through the web interface.
Thierry
Esteban
Thierry