* point jumping strategy

* pylint and flake

* extra param method for counting

* skip test with database that is to slow

Co-authored-by: Émile Nadeau <[email protected]>
Co-authored-by: Jay Pantone <[email protected]>

* requirement and row and col placements strategy pack

* f

* skip test with database that is to slow

Co-authored-by: Émile Nadeau <[email protected]>
Co-authored-by: Jay Pantone <[email protected]>

* limit the size of obs in minimal obs call

* if no obs, can't infer obs

* resurrecting old deflation strategy

* Add get equation method in the easy case

* add test for json/repr methods

* remove the skew/sum indec condition

* unused import

* add empty cell obs during placement

* CHANGELOG.md

* update partial placement test

* updating black version

* black formatting

…tegy (#425)

* a selective symmetry factory

* replace symmetries with this new class, default to all syms

* no basis given = all syms

* mypy

* changelog

Co-authored-by: Jay Pantone <[email protected]>
Co-authored-by: Émile Nadeau <[email protected]>

* an alternative sliding strategy based on a fusion argument

* add encoding tests

* if a column, then rotate -> slide -> rotate back"

Co-authored-by: Émile Nadeau <[email protected]>

* a targeted cell insertion strategy

* test the encoding

* remove double import

* circular import

* Update requirement_insertion.py

* disable import order warning

* actually also need to make the import comply to isort

* allow a list as input to strategy

Co-authored-by: Émile Nadeau <[email protected]>

* Actually adding the files this time

* Pleasing black

* tidying up

* update string method, test

* mypy

* formal step test

* Revert "formal step test"

This reverts commit fd41137.

* formal step test

Co-authored-by: Christian Bean <[email protected]>

* a point corroboration factory

* change to insert into any pair of cells which can not be mutually positive

* mypy

* remove positive corroboration from bijection packs

* remove unused import

* README.rst

* removing strategy from bijections test, and increasing timeout on forest test

* allows swapping points or assumptions alone

* fixing tests

* add flags

* adding init flags

* remove test for fused tiling when unfusable

* the cell reduction strategy

* manually change local basis to monotone in non finite cells, forbid if ob touches row and col

* add a timer to init to catch slow initialising

* add a heuristic to avoid large products

* change pattern placement to factor sooner

* an actual solution? Pick the product smarter.

* tidying up

* tidy up cleaned obs to reduce size of product

* point corrob

* cleaning

* tidying code

* remove point in bijections

* test reqs separately

* not

* allow reqs, and more careful point in between

* pylint

* api.combopal -> api.permpal

* make not implemented reverse constructor for interleaving factor

* update black version

* creates kitchen_sinkify function

* don't add short obs or obs inferral if the arguments passed are 0

* fixes CHANGELOG

* adds DB verified and req corrob

* tell pylint to go think about it more

* add test of kitchen sinkify pack json

Co-authored-by: Christian Bean <[email protected]>

Now that we do strategic garbage collection in some of our code some
test where timing out because the gc.collect() call where taking to
long. I introduced a new decorator to collect before those test to avoid
triggering out the timeout.

New collect_before decorator for some test.

* use cell decomposition for uniting cells in factor

* add component verification and add it to packs

* ignore component assumptions in rearrange

* more selective when verifying tilings with components

* flipping sum and skew in symmetry strategies

* if sum fusion, no skew assumption in region and vice versa

* checking assumptions in comp fusion fusable check

* not implemented counting for positive or empty fusion

* ensure make_tracked adds component strategies

* smarter is_component method to impove DetectComponentsStrategy

* get_minimum_value changed for comp ass since can have fewer components than points

* CHANGELOG.md

* allow comp in monotone and insertion verified

* component -> point assumption strategy

* verification test packs allow comps in insenc and monotone

* fix from dict

* fix to_jsonable

…448)

* remove AddInterleavingAssumptionsFactory

* allow untracked interleaving

* self.

…my strategy, and allowing directionless pp (#453)

* pointing strategy, a dummy strategy, and allowing directionless pp

* add json encoding test

* mypy and pylint

* CHANGELOG.md

* unfusion strategy

* factory

* the disjoint fusion strategy

* changelog and add to kitchen sink

* added counting code

* implement counting with params for unfusion and pointing

* pointing subtracts already placed cells

* reversible if parent has all the assumptions for interleaving

* unused import

* fix the param when generating objects

* remove empty cells when mapping assumption in interleaving factors

* remove outdated todo

* a strategy for placing points of assumptions

* make sure to do this on point tracking assumptions

* a factory for placing into rows and cols that are fusable.

* __str__

* test encoding

* adds to kitchen sink

Co-authored-by: Jay Pantone <[email protected]>

* add assumption reversible when assumption covers whole tiling

* check the assumptions being added

* add a test for is reversible on assumption insertion

* fix json and repr encoding

* remove dead code

* a requirement pointing strategy

* allow for directionless placement of gp into all cells, so divide by N

* only point at reqs on tilings

* add remove constructor

* fix equation

* remove unused import

* remove print

* first draft, mostly copying SubclassVerification.compute_subclasses

* remove prints and tidy

* update test

* Obs inf early yield (#472)

* adds yield_non_minimal flag to obs inf

* adds typing and sets to True

Co-authored-by: jaypantone <[email protected]>

…474)

* full-tiling assumptions do not count against the max_assumptions cap

* changelog

* adds flags to PointingStrategy and AssumptionPointingStrategy; also adds RequirementPointingStrategy to kitchen sink

* fixes reprs, from_dicts, to_jsonables

* fixes from_dict bugs from max_cell argument

* Bug fixes

* better __str__s

…#473)

* adds flag to allow factorable requirement insertions, default = False

* changelog and default set to True

* small bug in initialiser

* add a test

…#482)

* one by one looks up permpal

* remove cached property

* more detailed error messages

* reverse fusion uses divide by k

* mypy

* stop expanding av(012), and return invalid op for comp ver

* remove outdated test

* flake

Co-authored-by: Jay Pantone <[email protected]>

* change error to logging check

* Update test_reverse_fusion.py

* ensure obs are minimised in point placements

* changelog

* Work in process getting rid of the new warnings

* Updating black version in the pre-commit file

* fixing pylint

* Fixing typing errors

* Still fixing typing issues

* No more typing today, thank you

* A few more typing issues taken care of

* A bit of linting

* Pylint is now happy

* Mypy is now happy

* Quick fix for typing error of array

* "pypy3" -> "pypy3.8"

* change default from cpython 3.8 to 3.10 and add pypy3.9

* add quotes

* add more quotes

* MOAR QUOTES

* oops too many quotes

* add consecutive value condition for sliding

* tidy up

* check in decomp func of sliding

* changelog

* a new pack

* add json encoding test

* a first draft of a tracked classdb

* make default for guided and limited searcher

* adds flags to PointingStrategy and AssumptionPointingStrategy; also adds RequirementPointingStrategy to kitchen sink

* fixes reprs, from_dicts, to_jsonables

* fixes from_dict bugs from max_cell argument

* Bug fixes

* lots of improvements to trackedclassdb

* mypy

* mypy

* mypy

* black

* css now takes classdb flag

* use classqueue flag

* fixes the fact that assumption types were completely ignored

* adds one test

* removes some comments

* ignore pylint error

Co-authored-by: Jay Pantone <[email protected]>

* one by one uses permpal for counting etc

* fixed small bug getting rational gfs

* changelog

* only storing 1x1 specs without ass and reqs

* ensure not over correcting for reqs

* simplfy req correction line

* eqs print F_Av(basis)

* fixes bug in sanity checking new 1x1 verification

* reuse variable

* implement positive reverse fusion counting

* If left only and right is positive, then add one to every left parameter

* tidy up tests

* a test for duplicated mapped params in complement

* skip test for incr with assumption

* a strategy for forcing assumptions to be empty

* add relax assumption factory to kitchen sink

* tidy up kitchen sink method

---------

Co-authored-by: jaypantone <[email protected]>

* remove break

* changelog

* adds kitchen sink levels

* change log and tes

* the shift fix discussed on 16/05/23

* changelog

* fix imports