Re: Database Designer report at 28/07/2001

On Fri, Jul 29, 2011 at 4:34 AM, Dave Page <dpage(at)pgadmin(dot)org> wrote:

> Hi Luis
>
> On Fri, Jul 29, 2011 at 2:59 AM, Luis Ochoa <ziul1979(at)gmail(dot)com> wrote:
> > Hi, I don't report to this list (just to gsoc one) from some time, mostly
> > because I don't want to spam it.
>
> You should do - there's less chance of us rejecting changes if you
> keep us up to date :-)
>

Ok, I'm going to send any important update to this list for community review
and suggestions.

> > But, right now I want to share the status at the database modeler as
> today.
> > As you can notice (and test at pgAdmin repository) right now you are able
> to
> > create and save / load a single model in the database designer. A lot of
> > bugs have been fixed, and we have now a more stable designer but still is
> a
> > work in progress that will change in some time.
>
> I just gave it a whirl for the first time (using the code committed to
> master, so some of these may be fixed), and immediately ran into a few
> issues:
>
> - Generating a short table name from a long one caused an assertion.
> See the attached screenshot (what's the point of that feature anyway?)
>

Short Name is a feature used by auto-naming foreign keys subsystem, isn't
mandatory to use it, but if you want more readable names generate in an
automatically way for foreign key constraints is good idea to use it. This
feature is on testing now, if people like it this feature stays if not it
should be suppressed.

> - The table model is so small it's unreadable.
>

Table figure is generated using as basis the font size, if user change
default font size for the modeler tables automagically change it. I'll add
at next patch a menu item to allow changing the default font and size of
modeler.

> - Another assertion, when trying to add a column.
>

I'll check it, this is because all menus at figures are generated
automatically from a list of strings inside the figure and theirs event ids
are automatically assigned and as I can see in the screenshot I probably use
at some point 0 as the base and then this id was used for first item. Mostly
because I don't test it on a mac ;) I'll fix it for next patch.

> - The generated SQL lacks quoting of identifiers.
>

SQL generation subsystem will change for next patch, because I want to
generate tables first and later add constraints to avoid dependencies and
when I do that I'm going to add this suggestion. Just to clarify when you
said quoting of indentifiers we are talking about quoting tables names and
columns names?

- The generated SQL isn't syntax highlighted.
>

Well, right now this wasn't on my schedule but I'll try to do that at last
week of proposal dedicated to bug hunting and improvement of UI if time is
enough.

> - We'll also need to work on the icons - I'll probably be able to help
> with that.
>

Yes, right now any help with icons will be great because sometimes icons are
just a stub or borrow from another feature.

> The other issue I ran into is one of design - when adding a new
> column, we should ask for all the properties of the column on a
> dialogue at once (can you reuse dlgColumn, or at least
> dlgColumn.xrc?). Specifying just the name, and then having to go and
> manually adjust the rest of the properties is really quite tedious and
> not a good UI design.
>
>
Yes, that is an interesting and needed feature that I have been thinking
about (almost from begining), but is a little complex (because way columns
are stored inside the graphical model) to do it right now for next patch,
I'll try to do it at the UI improvement week at the end of schedule or later
but outside gsoc.

> > This week and last two I have been working on multiple views for same
> model
> > feature:
> >
> > This feature will bring some improvements over old one model approach for
> > the designer, a slides presentation with this feature can be shown online
> at
> > google docs:
> >
> > http://goo.gl/XT2Az
>
> Looks interesting... of more interest to me though is reverse
> engineering of an existing schema. Is that on your roadmap?
>

Well, in the proposal I have some kind of reverse engineering of tables
features on my schedule, called table sync because this feature will allow
to add tables from a schema to the model and then generate a alter table
instead of a create table for this imported table. But only if model is
generated to be created at same schema, in other way it generates a create
table.

This feature allow to choose several tables from same scheme at same time
and in this way import relationships between selected tables too. But right
now I haven't any design for it (at the middle of this week after next patch
I'll start to work on it), and I have not a real idea about how this feature
will really work.

> Right now I still continue working on the feature and so far I have
> > implemented a new approach for my hotdraw library that allow me to show
> in
> > different views the same object, in a first step this was done for normal
> > graphical objects like squares circles and so far, later it was done for
> > polygon figures (harder case because each view can have different number
> of
> > segments and other things). Later all database designer figures were
> > adapted to this new approach, and
> > right now I'm working in a way of reusing created objects inside the main
> > model in all diagrams (a treeview with that info).
>
> OK. Does that include the renaming we discussed?
>

Yes next patch will rename wxhotdraw to hotdraw, and prefix wxhd to hd.

> I hope to finish this feature by weekend or middle of next week, it
depends
> of bug fixing because it have been a very deeply change in the library,
that
> don't allow in their native architecture to show same figure in different
> views at same time.

OK.
>

> > At the presentation is included a screenshot at the end with the proposal
> > UI.
>
> s/A probably solution/A probable solution
>

:$ sometimes I'm tired and my english that is bad, becomes worst :(

Thanks for your comments Dave

Regards, Luis.