Chores #445
Chores #445
Conversation
```
Graphics/Implicit/Definitions.hs:113:2: warning: [-Wtype-equality-requires-operators]
The use of ‘~’ without TypeOperators
will become an error in a future GHC release.
Suggested fix: Perhaps you intended to use TypeOperators
|
113 | $(makeWrapped ''ℝ3')
```
sorki
force-pushed
the
srk/chores
branch
2 times, most recently
from
05dbfa5 to
5aa2c43
Compare
There are some new warnings about specialisation of imported functions
like
```
programs/docgen.hs: error: [-Wall-missed-specialisations, -Werror=all-missed-specialisations]
Could not specialise imported function ‘ghc-prim:GHC.Classes.$fEq[]_$c==’
Probable fix: add INLINABLE pragma on ‘ghc-prim:GHC.Classes.$fEq[]_$c==’
```
which we can't do much about. Don't fail in CI on these.
This fixes show for rotate3
so cheaper `r /= 0` is evaluated first and it can sometimes fail faster without evaluating more expensive `abs (x-y) < r` operation.
Before, only `Insidedness` was rendered, now we also show some more
details
```
3) symbolic obj 3, inverses, scale inverse
Falsified (after 1 test and 1 shrink):
V3 0.0 0.0 3.328273472768273e-2
(sphere 0.0,(V3 0.0 0.0 0.0,()))
TestResult:
| Outside
| scale (V3 NaN NaN 1.0) (sphere 0.0)
| Sampled at V3 0.0 0.0 0.0 returns NaN
/=
TestResult:
| Surface
| sphere 0.0
| Sampled at V3 0.0 0.0 0.0 returns 0.0
```
so `NaN` doesn't result in `Outside`. We also make sure that two NaNFails are never equal, so the property producing NaN always fails.
this (introduced by me) looks like a good idea (and identity) but it isn't since sometimes we want to reset rounding to zero for nested objects. This broke `prop "withRounding always takes the last value idempotent"`.
| @@ -56,7 +56,7 @@ getImplicitShared ctx (UnionR _ []) = | |||
| getImplicitShared ctx (UnionR r symbObjs) = \p -> | |||
| rminimum r $ fmap (flip (getImplicit' ctx) p) symbObjs | |||
| getImplicitShared ctx (IntersectR _ []) = | |||
| getImplicitShared @obj ctx Full | |||
| getImplicitShared @obj ctx Empty | |||
There was a problem hiding this comment.
I think so, based on https://en.wikipedia.org/wiki/Intersection_(set_theory)#Algebraic_properties
The intersection of any set with the empty set results in the empty set;
There was a problem hiding this comment.
Also our implicit space (or starting object?) seems to be emptySpace - if we have union [oneObject] or just oneObject (without implicit union xD) it seems like it can only be unioned with emptySpace, because if it was a fullSpace we would get fullSpace instead of oneObject.
|
|
||
| ------------------------------------------------------------------------------ | ||
| -- | The number of decimal points we need to agree to assume two 'Double's are | ||
| -- equal. |
There was a problem hiding this comment.
excuse me while i hurl. :)
There was a problem hiding this comment.
I think that the reason it wasn't required was that Arbitrary Double / Gen Double wasn't generating small enough or close to zero things and transforms. Before, there was epsilon = 5 and currently even epsilon = 16 works quite well but eventually, with large enough --qc-max-success it breaks. This seems to hold up to --qc-max-success=10M
Related to Haskell-Things#441 This will get factored out into canonicalization pass
It doens't hold for cases where one scaling has zero component. Adjust QC tests to generate scaling vectors without zero components.
|
I still want to factor out guards into canonicalization function, that will also solve it for |
sorki
force-pushed
the
srk/chores
branch
5 times, most recently
from
e4c04eb to
7c647c6
Compare
Implements recursive tree rewriting to handle identities, transformation merging, sanitization and basic handling of degenerate objects.
It didn't remove temporary files on successful runs, filling up tmp. Make `goldenFormat'` higher order so it can be used by both 2D and 3D. Render pretty fail message.
| -- Copyright 2015 2016, Mike MacHenry (mike.machenry@gmail.com) | ||
| -- Released under the GNU AGPLV3+, see LICENSE | ||
|
|
||
| -- | This module implements canonicalization pass that |
There was a problem hiding this comment.
| -- | This module implements canonicalization pass that | |
| -- | This module implements a canonicalization pass that |
| @@ -112,7 +112,7 @@ buildShared (WithRounding r obj) | r == 0 = build obj | |||
|
|
|||
| buildShared(UnionR _ _) = error "cannot provide roundness when exporting openscad; unsupported in target format." | |||
| buildShared(IntersectR _ _) = error "cannot provide roundness when exporting openscad; unsupported in target format." | |||
| buildShared(DifferenceR _ _ _) = error "cannot provide roundness when exporting openscad; unsupported in target format." | |||
| buildShared(DifferenceR {}) = error "cannot provide roundness when exporting openscad; unsupported in target format." | |||
There was a problem hiding this comment.
why quash the argument count just here?
There was a problem hiding this comment.
Hlint complains about it when there are 3 and more
| @@ -3,10 +3,6 @@ | |||
| -- Copyright (C) 2016, Julia Longtin (julial@turinglace.com) | |||
| -- Released under the GNU AGPLV3+, see LICENSE | |||
|
|
|||
| -- allow us to specify what package to import what module from. | |||
| -- We don't actually care, but when we compile our haskell examples, we do. | |||
| {-# LANGUAGE PackageImports #-} | |||
There was a problem hiding this comment.
No idea! (found by hlint).
Arbitrary Doublegenerator that now also generates exact zeroes.