Introduce and use PiceTrait downcaste_piece extension methoc

This is more convenient.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

gamestate: Add divider comment

We're going to add more above here, and it's getting big.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fastsplit: Move Z check from multigrab into fastsplit impl

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fastsplit: Correct z order

This was back to front.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Reorganise Z setting into two phases

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Move op_do_set_z to gamestate

This is preparatory, to reduce diff noise.  We're going to change the
way this works in a moment.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fastsplit: Move a comment to its right place

Some time in the development of this, pre-squash, kthe modperm was
acquired within the infallible closure.  But when it was moved,
evidently the comment didn't move with it.  Fix that.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fastsplit: Fix recovery

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

currency: Fix and improve multigrab (splitting) log message

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

multigrab/fastsplit: Plumb ShowUnocculted and &PieceTrait

The implementation doesn't really want PieceRenderingInstructions,
just permission to get at trait objects.

The currency fastsplit implementation callback is going to want to
look at its Banknote so that it can make a nice log message.  I think
fastsplit piece impls are entitled access to their concret piece type.
So downcast it, and pass it down the layers.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

PieceTrait: impl Downcast

We're going to want this in a moment.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

currency: Many fixes to splitting

This was quite broken.

* Actually set the value of the new (change) piece.
* Set the value of tghe taken piece correctly.
* Set the WRC so that clients see the taken piece value.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

currency: Move some assignments about

Tidying up.  No functional change.
(Was originally part of the next, bugfix, commit.)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

currency: Rename value

Having the two pieces here is confusing.  All the variables should say
which piece they are.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

currency: wip test

So far we test that splitting appears to succeed, but not what the
results look like.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

currency: Implement op_multigrab and therefore splitting

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
ng

fastsplit: Implement splitting

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

currency: Make currency pieces be fastsplit

The actual machinery does all the hard work.

This shouldn't have any overall visible functional change.  It just
engages new contraptions during piece creation and load/save, but
right now they shouldn't have any work to do because we don't actually
ever split (or merge) currency.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fastsplit: Add some imprecations about not dropping FastSplitId

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fastsplit: Introduce fastsplit concept

This has all the relevant data structures and support code, but does
not yet implemnet the actual splitting and deleting/merging
operations which will be needed for the piece api op impls.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Drop an unneeded lifetime argument

This is not needed.  (I'm not sure whether I did this because
ambassador didn't like the lifetime argument, or just because I saw it
in passing.)

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

ambassador: Introduce dependency and make some traits delegatiable

fastsplit pieces are going to want to delegate everything.  Not only
is this a lot of typing, open-coding it risks bugs when new trait
methods are added.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

multigrab: Set the z coordinate

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Break out op_do_set_z

We're going to make multigrab do z setting too.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

multigrab: Check that the piece is not held

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Have OpOutcomeThunk::Deferrecd take &mut InstanceGuard

That way it can call modify_pieces_not_necessarily_saving_aux, which
the fastsplit_split implementaion wants.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

api: setz: Avoid an unwrap

Panicking is worse than failing.  Spotted this in passing.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

ModifyingPieces: Provide on-your-own-head constructors, and doc

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Document z coordinate sense convention

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

script: multigrab: swap two if arms

This seems clearer, by putting the usual case first.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

script: piece_raise: allow overriding the actual implementation

We're going to need to do something else for multigrab.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Change type of UnpreparedPreparedUpdates

We're going to want to compose this in an additional place.  And all
this Option and None is awkward.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Introduce and use PrepareUpdatesBuffer::only_unprepared_with

This saves us writing `if unprepared.len() != 0`.  I think that
pattern is asking for a reversed conditional at one random site.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Introduce and use methods for applying UnpreparedUpdates

There are quite a few call sites where the nature of an
UnpreparedUpdates is open-coded.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Use default() for empty UnpreparedUpdates, not None

We are going to make this into a vec![].  Doing this change now will
reduce the textual impact a lot.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Drop use of fehler in PieceTrait definition, for ambassador

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

occult ilks: Provide clone_id and clone_iilk

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Introduce OpOutcomeThunk

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Introduce PieceLoadArgs

So that we can add more parameters to PieceTrait::load() without
having to edit everywhere *again*.

In fact it turns out I don't need to do this right now (for currency
and fastsplit) but I think it's worth keeping.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

PieceUpdateOps: Tabular formattint

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Introduce non-Quiet PUO::Insert

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

ts: Break out insert_piece function.

We'll reuse this for non-Quiet Insert.  Have it return p for
convenience.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Rename PUO::Insert to InsertQuiet

This doesn't produce a halo, so it's Quiet.  We're going to want one
which *does* produce a halo, so we must make room.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Introduce and use xdata_init

Using get_mut for this seems slightly less careful.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Document what PUO::*Quiet mean

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Document OccultOwningId serialisation and refcount rules

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fix serde_with_compat attribute syntax

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Switch to Rust 2021

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

PROTOCOL.md: fix a typo

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

hidden: Add checks for un-notched to vpid::consistency_check

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fake rng: Remove an Arc

We don't in fact need this to be Clone, now that ServerConfig isn't.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

at-hidden: Add some more synch, just to complete things

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

at-hidden: Add more test, completing repro of unnotched bug

This test fails without the unnotched piece tracking.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

hidden: Fix handling of unnotched (Distinct) pieces

We need to un-occult them.  But we were tracking what to recompute by
looking in occ.notches.  Un-notched pieces aren't there.

Introduce a new HashSet for these.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

hidden: Refactor a bit in preparation for unnotched pieces fix

In both of these places, we're going to want to do something if there
is an occultation but notch is None.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

hidden: Have active_total_ppieces return usize

The fix for the un-notched pieces bug is going to want to add a
HashMap len to this, and that's usize.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

hidden: Introduce Occultation::pieces

This is going to want to return un-notched pieces too.  Fix both sites
where we iterate over the notches.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

at-hidden: Engage fake time feature

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

tests: Enable the fake time feature

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fake time: Make fields pub that should be

We want this for FakeTimeSpec in a test case.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fake time: Fix config deserialisation

Option works weirdly, empirically.  This (via a Vec) is what we want.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

dice: Honour fake time feature

Really we should make this more pervasive, but this is what we need
right now.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

fake time: Introduce new feature

Needs a little work still, and not used.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Make ServerConfig not be Clone

We're going to put GlobalClock in it.  It doesn't actually want to be
Clone since it's a singleton.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

tests: at-hidden: wip test

This currently passes, and therefore didn't repro the un-notched piece
bug which I have been trying to repro and fix.  But it's a start.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

logging: In api_piece_op, log the form too

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

tests: Increase debug traceing of hidden and updates

I wanted this for debugging the unnotched piece crash.  It seems fine
to keep it.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

dice: Allow loading old savefiles with no desc field

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

apitest: Break out filter_by_desc_glob etc.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

demo game: Give the dice descriptions

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

dice: Support specifying the html description in the spec

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

dice: Refactor describe_html to prep for desc

Break out showing so that we don't have combinatorical description
formats.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

demo game: Make the 2-sided die a disc

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Fix many dbgc calls to be trace_dbg instead

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Fix trace_dbg to tolerate call without any value parameters

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Move logging::setup out of main.rs

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

logging: Change output formats

Format is like that from flexi_logger but with only the level
coloured.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

currency: Move qty into xdata

When we multigrab, we must split, which involves mutating the value of
the grabbed piece.  So the value must be in xdata, not ipc.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

at-otter: Fix a ref to PIA that ought to be PIB.

I think.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

currency: Abolish min_unit

We can instead just rely on the players to Not Do That, or to tell
each other off.  It's easy to recover from, too: just move money until
the amounts are right.

This also avoids any question of what to do if people try to combine
banknotes with the same currency but different min_qty.  (GCD?)

Suggested-by: Gareth Taylor
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Multigrab: Provide in initial load

This was broken by the special refactoring.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Multigrab: allow to multigrab occulters

I don't expect any pieces to exist that can both be active occulter,
and multigrab.  But if it does it should probably do the thing, rather
than mysteriously not work.  And this simplifies the code.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

JS: Add a comment about PieceInfo vs SessionPieceLoadJson

I made a bug due to not realising this needed to be updated too.

Maybe at some point the common bits should be combined, at least on
the Rust side.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Currency: Declare to be multigrab

We don't actually implement the PieceTrait method, so that just
returns Inapplicable.  Ie, after this you can try to multiselect a
banknote but you get
  Problem manipulating piece: UI operation not valid in the curret piece state
in the log.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

JS: Implement multigrab

Check the multigrab boolean, and if appropriate make a single
multigrab API call rather than trying to grab multiple pieces.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Plumbing for "multigrab" piece API call

We can now pass this through to a piece.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Elide angle in game save file, when it's just North

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Elide some fields from save game state where we can

This saves writing out the field names as well as just the vslues.

We would like to do this for the updates sent to the client, but that
would really depend on the JS nullable coalesce operator "??" which is
not available the version of typescript I'm using, nor our ES
compatibility target.  Newer typescript would do a polyfill but that
maybe wouldn't be super-fast?  So postpone that.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Fix type of PreparedPieceState.held in the JS

I was comparing the JS and Rust versions, and spotted this
discrepancy.

In fact it is assigned to internal fields and variables which are all
nullable, so there is no other code change needed.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Provide is_default() function

We're going to use this to elide some fields from the game save state.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Provide and plumb multigtab piece property from Rust to JS

Currently nothing sets this (on the Rust side) nor uses it (on the JS
side).

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

script: Add a missing comma in a type definition

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Add "p" argument to mouse_clicked_one

Almost the callers have it and the one that doesn't (mouse_find_clicked)
is about to need it in another of its branches.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Move special rendering instructions to PieceSpecialProperties

This gets rid of an ipc method (meaning we get rid of one method call
during piece rendering) and is also more principled, than a call which
might conceivably give different answers at different times.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

Add special field to SpecLoaded

This is currently empty.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

split PieceSpecLoaded

We are goint to want to add a field to only the non-inert version, so
this generic is not applicable any more.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

api: Introduce and use ApiPieceOpArgs::pri

We're going to have a new call site, too.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

api: Replace open-coded calculation with call to fully_visible

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

ui: Disable pointer events on text in dice and currency

Otherwise you can end up with bits of the displayed text (currency
value, or die face) selected.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>

ui: Break out HTML_TEXT_LABEL_ELEM_START from clock.rs

We actually need this in other places too.

Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>