shapelib: Substitute _colour multiple times if necessary

No, or many, is not an error here.

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

shapelib subst: Provide subst_general and change substn

We're going to want this convenience alias.  substn's previous
signature was not helpful.

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

shapelib subst: Provide subst arg to colour_subst_1

We're going to want to call substn too.

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

shapelib subst: Provide substn

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

shapelib subst: Add test cases for subst()

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

shapelib: magic: Actually substitute _colour in the template

This doesn't actually work because _colour appears multiple times and
subst insists on it being there only once.

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

shapelib: Add a trace call for the magic template

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

otterlib: Initialise the nwtemplates

Otherwise library items that depend on magical pieces that use these
templates (eg, dice) don't work correctly.

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

nwtemplates: Split init_from_config and init_from_dir

otterlib is going to call init_from_dir.

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

nwtemplates: Log a warning if there are no templates

This is probably an installation or usage error.  Reporting that helps
if subsequently a template isn't found.

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

otterlib: Pass opts to preview()

This is going to want the nwtemplates directory.

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

otterlib: Plumb gpc through from load to render

Some magical pieces require their xdata (which they set in load) for
rendering.  And now we can have library pieces which contain magical
pieces which otterlib must therefore handle.

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

Makefile: Break out NWTEMPLATEASSETS variable

We're going to want this for the dependencies of the shapelib,
since otterlib will need to read this template.

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

shapelib magic: implement

This still has some bugs.  We mustn't invoke it just yet in our
library bundles, because it will fail.

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

spec: LibraryItemNotFound: Actually report the ItemSpec

This error message was hopeless.

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

Make TomlQuote pub

Like it should have been.

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

shapelib: Break out subst_item_name

Magic is going to want this.

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

shapelib: Break out format_item_name

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

shapelib: Rework colour substitution

There were too many repetitions of the variable names etc.

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

shapelib: Fix sort key colour abbrev

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

Provide TomlQuote instead

This is slightly better in terms of consing.

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

Provide toml_quote_string

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

Add lazy_regex dependency

Sadly we can't use this in spec.rs because it's not compatible with
concat!

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

shapelib: Break out Catalouge::add_item

We're going to want to call this again in add1, I think.

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

shapelib: Break out process_files_entry

load_catalogue was getting quite unwieldy.

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

shapelib: Implement Magic catalogue entries

But nothing constructs them.

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

shapelib: Remove now-needless type annotation

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

shapelib: Introduce CatalogueEntry enum

Currently it only has one variant, but this gets us the plumbing.

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

shapelib: Split find_then from find_load

This is preparatory to making the catalogue capable of holding magic
pieces.

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

shapelib catalogue: Get description from trait

We need to disambiguate this right now because we have a concrete
item here which impl both PieceTrait and InertPieceTrait.

That's going to change later.

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

shapelib catalogue: Get sortkey from trait

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

shapelib catalogue: Make ItemSpecLoaded and find_load private

We're going to make this more complicated.  Let's make it private and
have otterlib use the trait methods.

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

docs: Correct wrong bundle version

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

dice: d6: Tiny tweaks

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

dice: d6: image: Add to edited.toml and demo.toml

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

dice: d6: image: Put comment at the end

usvg seems to hate it at the start.

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

dice: d6: image: Rename output file to match intended recolouring

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

dice: d6: image: wip extraction

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

dice: d6: Add label to the die part of die.svg

This will let us extract it for use in a shape library.

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

shapelib: Rename OutlineDefnTrait

This is clearer as we have the Enum too.

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

mformat 2: Actually implement the outline resizing

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

shapelib: Fix an improperly quoted item in a doc comment

Without the quotes, rustdoc wants this to be a hyperlink.

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

mformat 2: Actually parse the outline properly

This turns out to be quite hard.

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

ambassador: impl_via_ambassador: Make it handle multiple

The result is quite nice.

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

ambassador: impl_via_ambassador: Put back the type name

IDK what ambassador's body_struct etc. macros might use this for in
the future.  Keeping to the pattern that was in response to my MR
seems wisest ?

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

ambassador: impl_via_ambassador: Introduce and use it once

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

ambassador: Remove some empirically-unnecessary arguments etc.

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

ambassador: Switch to upstream version of delegation to fn

Directly as per:
  https://github.com/hobofan/ambassador/pull/32#issuecomment-1121759145

Quite ugly, I think.  I will try to tidy it up somehow.

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

docs: Fix two Sphinx syntax errors

These just cause misrendering, at least in my version, not an error
message.

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

Document that we need Nightly Rust for the full tests

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

mformat 2: Declare provided table and game specs format=2

This has no functional change, but it sets a good example.

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

mformat 2: Insist that builtin libraries have an explicit version

We don't want to be setting the bad example of a missing `format`, and
of course we don't want to be using format=1 either.

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

mformat 2: Convert all builtin libraries

This mostly involves deleting (enables us to delete) various size
specifications which can be inferred from the SVGs.

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

MaterialsFormatVersionError: Improve messages

Well, I managed to produce these.  But the output was a bit ugly.

I think the stuff in Other about a "reasonable integer" is covered by
the specific strings we pass in the integer conversions, and there's
going to be a possibility of completely different kinds of problem:

Use of a format default in a builtin shape library is going to become
an error.  And we don't want to add a variant for it - that's
overkill.  So we'll reuse Other, and this message is not right.

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

mformat 2: Implement the new size handling

As documented.  This is a much less confusing approach and doesn't
involve recapitulating the SVG's size, which Otter now knows.

It is easier to provide a complete reimplementation than to refactor
and reuse bits of the old.

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

shapelib: Introduce and use LLE::InGroup

Otherwise bundle errors are really quite inscrutable.

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

shapelib: Introduce and use LLE::InLibrary

Otherwise bundle errors are really quite inscrutable.

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

impl Mean for f64

We'll want this to calculate middles of SVGs.

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

mformat 1: Drop the size parameter from load_mf1

This was introduced in
  size handling: Plumb svg size through to OutlineDefn
and it seemed a good idea at the time.  But when I wrote that I hadn't
decided that these entrypoints would be renamed and kept, rather than
start to contain conditionals.

The mf1 versions can't depend on the size, since in format 1 the SVG's
actual size wasn't used.

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

mformat 2: Provide OutlineDefn::load which takes an explicit size

For circles we no longer insist on width and height being equal (in
fact, insist on only being provided with one number), as we did for
mformat 1.  Rather we use the maximum of the two dimensions, for both.
As documented.

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

shapelib: outlines: Fold get* into load_mf1

Now that we are doing checking by loading with a dummy size, and the
separate check methods are gone, the get_mf1 and get_size functions
each only have one call site.

Removing them makes the code simpler.

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

mformat 2: Have resolve_square_size return Option

This moves the optional-ness handling into resolve_square_size, and
pushes the error into the call sites.

The mformat 2 callers are going to want to default this, so want an
Option.

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

mformat 2: Actually tolerate missing size during deser

We don't parse this properly yet, but we will.  This bug was found in
testing (with commits further down the branch) but I have rebased it
next to the code that introduces the bug.

Also explain that the weird scaling in GroupData only happens in mf1.

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

mformat 2: Deserialise the data that library format 2 needs

Introduce the various enums (including serde(untagged) ones) so that
we can deserialise all of the mf1 and mf2 formats into the same
structure.

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

mformat 2: Allow materials_format::Version to contain 2

Nothing changes in the behaviour yet - nothing checks this - so right
now, this format is not functioning as documented.

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

mformat 1: Have FaceTransform::from_group_mf1 take the GroupData

This gives access to mformat, which is useful for error
reporting (since it allows us to use the standard "can't do that in
this mf version" constructor).

Of course this is an _mf1 function so it doesn't need the actual
value - it could perhaps have made do with a proof token of const
generic type, or something.  But I tried that and it was a palaver.

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

mformat 2: Rename various functions to have _mf1

These entrypoints are going to be specific to format 1.

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

mformat 2: Document new format

This is not yet implemented.  Documenting it first will make these
next commits have some sensible context.

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

mformat: Provide Incompat, utility functions, LLE variant, etc.

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

shapelib: Rename nominal arg to rect from centre

This is the nominal location, around which the rectangle is
centred.  This makes it clearer exp. for the call sites elsewhere,
where this parameter is just the piece position.

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

shapelib: Implement check() as load() with dummy size

We really want to know that we will be able to call load().  The
purpose is early detection of errors.

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

Fix BuiltinLibrary::default_materials_format

This needs to be 1, for now, since all our libraries are still in
format 1.

This was wrong since it was introduced in
  materials format: Plumb through to load_catalogue
but hasn't mattered because nothing in the shapelib code actually
looks at the format yet.

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

Makefile: Always run miri on nightly

It's not stable.

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

shapelib: Break out resolve_square_size

We're going to want to reuse this even more.

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

shapelib: Reorganise shape/outline handling (2)

Centralise this in two functiosn now associated with GroupData.

Now there is one place we can put new size and outline handling.

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

shapelib: Reorganise shape/outline handling (1)

Rename OutlineCalculable to ShapeCalculable.

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

shapelib::RectShape: Add a note

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

shapelib: Reorder file and introduce sections

This was getting quite hard to navigate.

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

shapelib: Promote call to group.d.outline.check()

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

geometry: Properly publish pos_zip_try_map etc., and use a bit more

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

geometry: Rename pos_zip_try_map, add infallible pos_zip_map, use

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

geometry: Introduce some affordances, and add a use case for each

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

docs: Document format compatibility

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

docs: Make room for format version in doc of bundle metadata

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

docs: Fix formatting of sub-entries in gamespec

Re-using the same bullet point symbol confuses Sphinx.

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

Makefile, make-release: Make published version current

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

Makefile: mkdir the version-specific sphinx output dir

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

Makefile: Break out PUBLISH_DOC_SPHINX_BASE and _TAIL

We're going to want this so we can mkdir.

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

Makefile, make-release: Publish to unreleased/docs/ by default

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

docs: Fix wrong information about builtin shape libraries

shapelib-toml.rs is basically obsolete now - we just haven't
dismantled it yet.  Now this is all in docs/shapelibs.rst.

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

docs: Fix an editing error in dev.md

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

shapelib: Plumb mformat through to OutlineDefn via GroupData

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

shapelib: Make fields of GroupData private

This was odd.  It compiles with them private, and it's not serde.
I want to add mformat here, which ought not to be pub here I think?
Anyway, I think this pubness was a mistake.

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

materials format: Read in game specs, but currently ignored

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

materials format: Plumb through to load_catalogue

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

materials format: Make constants be of type VERSION

This lets them be more easily used in constructors etc.

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

materials format: Provide try_from_integer

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

materials format: String parsing

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

materials format: Introduce version type

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

main lib.rs: Sort the hyphen imports

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