On Mon, Mar 2, 2015 at 12:34 PM, Tudor Girba <tudor(a)tudorgirba.com> wrote:
Hi,
On Mon, Mar 2, 2015 at 9:08 AM, stepharo <stepharo(a)free.fr> wrote:
Guys
Do not do that. You can load them in one click.
Moose is already bloated.
Each time I try to make moose reloading correctly
I have to fight with dependencies badly
expressed.
So do I. But, that is because we do not have proper tool support for
sanitizing configurations. Seeing the issue when we have small and modular
dependencies is a symptom, not a cause. We have to build proper analysis
tools. I am actually upset at myself for not having done that already. I
started with the support in the inspector for easier browsing, but we need
more.
First, I would want to try to achieve to move all packages from
ConfigurationOfMoose into Units and make ConfigurationOfMoose be an
Assembly. Then, I would like to reach the state in which one command to be
able to trigger a full Moose release including nested configurations. Once
we have that, we are in a manageable situation.
Alex if you need CSV and JSON just load them.
It's not just Alex. I needed that quite regularly, too, but that does not
mean anything either. I would rather want to look at the problem from a
principle point of view.
Moose is a distribution for data analysis, and having support for basic
data formats is a basic requirement for such a platform. There are other
components that are less needed, such as Magritte (which is used only in
Metanool and could be made simpler) or even the PharoImporter which is
likely not used in many places.
I do agree that Moose can slim down, but keep this in mind: the lines of
code of Moose remained basically constant since 4 years (around 180k) even
though the functionality increased dramatically. That is a good sign, and I
think we can do more to keep things slim but powerful.
The problem is that you prefer to let us manage dependency else.
> I really do not like that.
> I wrote a full chapter and it is working. To create a configuration it
> takes less than 5 min.
> So if you have a project just create a configuration and you get done.
> I really think that going that road
companies like synectique will
> simply
> have to fork (in fact they probably already have) and we will all lose.
Synectique chooses to play an opaque game, which is Ok. I have no access
either to the ways decisions are being made inside or even to the needs the
company has. And I won't spend much time guessing either.
Not entirely. We have been sharing info on the mailing list and the others
related to Pharo regarding different needs/impediments. The latest being
performance of PP and I wrote that we have to rely on PP v.1.51 because of
the performance issues with the latest version. I tried to load PP 1.51 in
Moose but reverting to an older version is not possible with Metacello (API
allows it but it wasn't actually happening because of some bug). Hence, I
had to fork the config of Moose to load PP 1.51 cleanly without any config
in the parts of Moose loading the latest version of PP. I can publish it
somewhere if people are interested.