[squeak-dev] The Trunk: Monticello-eem.617.mcz

Eliot Miranda eliot.miranda at gmail.com
Wed Sep 2 22:41:11 UTC 2015


Hi Tobias,

On Thu, Aug 27, 2015 at 10:56 PM, Tobias Pape <Das.Linux at gmx.de> wrote:

> Hi Eliot,
>
> On 28.08.2015, at 00:09, commits at source.squeak.org wrote:
>
> > Item was added:
> > + ----- Method: ChangeSorter>>monticelloChangeSetMenu: (in category
> '*Monticello-changeSet menu') -----
> > + monticelloChangeSetMenu: aMenu
> > +     <changeSetMenuShifted: false>
> > +     <menuPriority: 600>
> > +     "Sigh, when implementing menu pragmas this is not what I had in
> mind..."
> > +     aMenu add: 'delete Monticello load change sets' action:
> #deleteMonticelloChangeSets.
> > +     (aMenu submorphs
> > +                                             detect: [:m| m
> isMenuItemMorph and: [m contents beginsWith: 'destroy change set']]
> > +                                             ifNone: []) ifNotNil:
> > +             [:destroyItem| | item |
> > +             aMenu removeMorph: (item := aMenu submorphs last).
> > +             aMenu addMorph: item after: destroyItem].
> > +     ^aMenu!
>
> Can you explain what you want to achieve
> and how I can make this more convenient?
>

We can talk this over when we meet in Palo Alto, but the idea is an
important one, and I'm paying for having been too lazy to write it up
properly.  Let me try and give a quick sketch here.

The goals are to

- allow package load to augment menus without the original definitions of
those menus knowing anything about the subsequently loaded packages.
- to allow menus to be augmented with single actions defined in methods
that implement those actions
- and, depending on menu implementation, either
  - if menus are created every time a user invokes a menu, then to simply
include the relevant extensions in the menu
  - if menus are stored in variables, to automatically construct, or
reconstruct the menu when either loading or unloading of a package changes
the items in a menu

We design menu pragmas in action methods to specify
- the (name of the) menu to be extended
- the text of the menu entry
- information regarding what group of items the menu item belongs with
- information regarding the menu item's relative order with respect to
other elements in the group

Rationale:
With this specification a package can extend any number of menus in tools
simply by being loaded, with no collisions with any other packages defining
extensions on those menus.  Tools become freely pluggable, and system
configurations constructed from different sets of loaded packages are
decoupled; we don't need many menu definitions, just an initial base menu
which is extended with the current set of extensions.

For this to work we need to "name" menus, and to use some numbering scheme
to define groups (menu entries between lines) and elements within each
group.  One simple scheme is to use numbers:
- the definitions of initial base menus number groups, so that the first
group (the items before the first line) is numbered group 10, and the
second group 20, and so on, allowing extensions to specify new groups by
using group numbers that are not a multiple of 10
- the definitions of entries within groups are numbered, again in multiples
of 10, so that the extension can fit anywhere in the group

So given
MenuMorph fromArray: {
{'find...(f)' translated. #find}.
{'find again (g)' translated. #findAgain}.
{'set search string (h)' translated. #setSearchString}.
#-.
{'do again (j)' translated. #again}.
{'undo (z)' translated. #undo}.
#-.
{'copy (c)' translated. #copySelection}.
{'cut (x)' translated. #cut}.
{'paste (v)' translated. #paste}.
{'paste...' translated. #pasteRecent}.
#-.
{'set font... (k)' translated. #offerFontMenu}.
{'set style... (K)' translated. #changeStyle}.
{'set alignment...' translated. #chooseAlignment}.
"
#-.
{'more...' translated. #shiftedTextPaneMenuRequest}.
"
}

you could imagine the following numberings:

    10.01    find
    10.02    findAgain
    10.03    setSearchString
-
    20.01    again
    20.02    undo
-
    30.01    copySelection
    30.02    cut
    30.03    paste
    30.04    pasteRecent
-
    40.01    offerFontMenu
    40.02    changeStyle
    40.03    chooseAlignment

So I can specify e.g. a new group inserted before the first one by using,
e.g. 5.005 as an index, and can slot an item between again and undo using
20.015, etc.

This gives us menu pragmas that look like:

TextEditor methodsFor: '*SpellChecker-extensions' stamp: 'mad hacker
9/9/1999'
spellCheck
    <menuAction: #yellowButtonMenu
      label: 'spell check selection (s)'
      position: 10.035>
    ...code implementing spell checking...

(note that the action, #spellCheck, is implicit, it is the selector of the
method containing the pragma)

and
  - if menus are created every time a user invokes a menu, then search for
pragmas within the relevant class hierarchy, compute, sort and insert
individual extensions to the menu
  - if menus are stored in variables, to have compilation and method
removal send a message to the class to/from which method(s) are
added/removed so that the class can recompute menus (as in the line above)
when code is added/removed.

What I saw in ChangeSorter was quite different.  It seems to be a way of
specifying entire menus, and I'm not sure how these menus are to be
combined.  Maybe I'm misunderstanding the way the scheme is supposed to
work.  What I've described above is very close to the one that pragmas came
from in the first place, which was my attempt to decouple the components
that could be loaded into VisualWorks when we decomposed the system into
easily loadable parcels.  Up until that point VisualWorks had had all its
tool menus predefined, disabling items that were to do with components
delivered as file-ins that weren't yet filed-in.  It was too rigid.  It
didn't allow for future expansion.  And so I had the idea of menu pragmas
to decouple things.  Steve Dahl then came up with the scheme to define a
base menu, number its items appropriately, and extend the menu.

One of the things that's really nice about the VisualWorks scheme is that
the menu pragma itself is actually implemented (remember that pragmas are
Message instances with literal parameters).  The class MenuBuilder (IIRC)
implemented various menu pragma methods so the patten is

- in the method that defines the base menu, define the base menu, create an
editor on it, ask the editor to collect the pragmas and augment the menu
with those found
- the menu editor then collects the menu pragmas and then performs each one
to collect the information from it

So one can find other menu pragmas using senders, and find the editor using
implementors, and one can extend the scheme by adding additional methods to
MenuEditor, again possibly loaded from packages.


HTH

_,,,^..^,,,_
best, Eliot
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20150902/94b83351/attachment.htm


More information about the Squeak-dev mailing list