#162 - Add macros to automatically create boilerplate code #228

schillic · 2018-02-08T19:26:55Z

See #162.
Also resolves most of #145 (had to be done in one PR) (only ConvexHull missing).

schillic · 2018-02-08T19:29:36Z

Example: Let us define some nonsense: EmptySet is the neutral element of the LinearConstraint operation.

julia> l = LinearConstraint([1.], 1.); e = EmptySet{Float64}();

julia> @commutative_neutral(LinearConstraint, EmptySet);

julia> LinearConstraint(l,e)
LazySets.LinearConstraint{Float64}([1.0], 1.0)

julia> LinearConstraint(e,l)
LazySets.LinearConstraint{Float64}([1.0], 1.0)

julia> LinearConstraint(e,e)
LazySets.EmptySet{Float64}()

julia> LinearConstraint(l,l)
ERROR: MethodError: no method matching LazySets.LinearConstraint(::LazySets.LinearConstraint{Float64}, ::LazySets.LinearConstraint{Float64})

mforets · 2018-02-10T05:09:18Z

src/helper_functions.jl

@@ -1,6 +1,8 @@
 export sign_cadlag,
       check_method_implementation,
-       check_method_ambiguity_binary
+       check_method_ambiguity_binary,
+       @commutative_neutral,


i'm not sure that we want to export these names, since this is more of an internals of the package (same comment applies to the 2 check_... methods of above)

Good point.

But I think we need to export the check methods for usage in the unit tests (not tested). Somewhere I proposed that we remove them from the code base and move them to the unit tests. Any opinion?

even if it is the name is not exported, one can use the function by writing the module's name, as in LazySets.check_method_implementation, or it's used in many places do a import LazySets.check_method_implementation.

but i agree to move those check methods to the tests folder 👍

one can use the function by writing the module's name

You are right, I forgot that.

but i agree to move those check methods to the tests folder

Good, I changed #197 accordingly.

mforets · 2018-02-10T05:18:17Z

src/helper_functions.jl

+
+### Examples
+
+`@commutative_neutral(ConvexHull, EmptySet)` creates the following functions:


you can more synthetically write the equations, eg. "it creates the following functions for any set X: CH(X, ∅) = CH( ∅, X) = X and the edge caseCH(∅, ∅) = ∅ .

∅ might be misleading, but I get your point.

mforets · 2018-02-10T05:21:55Z

src/helper_functions.jl

+
+### Examples
+
+`@commutative_neutral_absorbing(MinkowskiSum, ZeroSet, EmptySet))` creates the


extra parentheses after EmptySet 👁

mforets · 2018-02-10T05:30:36Z

src/CartesianProduct.jl


- `CartesianProductArray()` -- constructor for an empty Cartesian product
+Constructors:


while we are at this line, a general thought is that i find the illustration of the different constructors to be more friendly if it is given as a proper example, rather than a copy paste of the source code -- if one wants that level of detail, then it's better to just look the code, by clicking on "view source", no? (or use Julia's @edit macro). that is to say that for me we can delete this part and instead put some examples of the different uses (in another PR).

I added #232.

mforets · 2018-02-10T05:40:52Z

src/CartesianProduct.jl

-function *(S::LazySet, cpa::CartesianProductArray)::CartesianProductArray
-    push!(cpa.array, S)
-    return cpa
+function CartesianProductArray(S::LazySet,


ok, this change is to save the symbol * for the binary operator only, in other words these two that used to be the same are now different:

julia> b = Ball2(zeros(2), 1.); julia> c = CartesianProductArray([b, b]) julia> CartesianProductArray(b, c) LazySets.CartesianProductArray{Float64,LazySets.Ball2{Float64}}(LazySets.Ball2{Float64}[LazySets.Ball2{Float64}([0.0, 0.0], 1.0), LazySets.Ball2{Float64}([0.0, 0.0], 1.0), LazySets.Ball2{Float64}([0.0, 0.0], 1.0)]) julia> b * c LazySets.CartesianProduct{Float64,LazySets.Ball2{Float64},LazySets.CartesianProductArray{Float64,LazySets.Ball2{Float64}}}(LazySets.Ball2{Float64}([0.0, 0.0], 1.0), LazySets.CartesianProductArray{Float64,Laz ySets.Ball2{Float64}}(LazySets.Ball2{Float64}[LazySets.Ball2{Float64}([0.0, 0.0], 1.0), LazySets.Ball2{Float64}([0.0, 0.0], 1.0), LazySets.Ball2{Float64}([0.0, 0.0], 1.0)]))

i doubt a bit, but maybe it is more sensible the new behavior.

Actually I wanted to discuss this.
Previously * and CartesianProduct were not identical. I found this inconsistent, because *(S, T) was said to be an alias for CartesianProduct(S, T).
Now it is identical, but we get what you observed.

I propose to additionally add a similar macro relating CartesianProduct and CartesianProductArray (and similarly for the other two) such that we also get the old behavior.
If you agree, should I do it here or in a follow-up PR?

follow-up PR

Okay, I took a note in the related issue #162.

mforets · 2018-02-10T05:57:34Z

src/helper_functions.jl

+* `MinkowskiSum(::ZeroSet{N}, Y::EmptySet{N}) where {N<:Real} = Y`
+* `MinkowskiSum(Y::EmptySet{N}, ::ZeroSet{N}) where {N<:Real} = Y`
+"""
+macro commutative_neutral_absorbing(SET, NEUT, ABS)


i was wondering if it's possible to work it around such that the extra work done by commutative_neutral_absorbing is done on the function commutative_absorbing, provided that the neutral element of the given set has been defined. we can do this by adding a function neutral(::set_type)::neutral_type and a check isdefined(neutral($SET)).

Hm, we have three of these cases, one with only neutral, one with only absorbing, and one with both neutral and absorbing. So I do not see why neutral should play a different role.

But I like the general idea. I would rather add two functions neutral and absorbing as follows.
The default implementation is neutral(set) = nothing and for the respective types we override with neutral(MinkowskiSum) = ZeroSet. Then we can get rid of two of the three macros. What would be the name of the remaining macro then?

Then we can get rid of two of the three macros.

how is that? i had in mind to substitute the call to commutative_neutral_absorbing to the calls commutative_neutral and commutative_absorbing. but one wants to do this such that the order of the calls doesn't matter :/

You just call this new function (call it fun) and internally it calls the neutral and absorbing functions to see what it should do. Does it make sense?

I understood now that what you proposed was very similar, sorry for the confusion.

with only 1 macro, the name commutative_neutral_absorbing is not the most intuitive IMO since EmptySet has nothing to do with _neutral_ of CartesianProduct, while we have to write:

# EmptySet is the absorbing element for CartesianProduct absorbing(::Type{CartesianProduct}) = EmptySet # create absorbing element functions @commutative_neutral_absorbing(CartesianProduct)

idea:

# EmptySet is the absorbing element for CartesianProduct @absorbing(CartesianProduct) = EmptySet

(or statement like: Emptyset @absorbs CartesianProduct or @is_absorbing_for)

that unrolls to absorbing(::Type{CartesianProduct}) = EmptySet and the function definitions macro. but for a case like M-sum one does:

@absorbing(MinkowskiSum) = EmptySet @neutral(MinkowskiSum) = ZeroSet

I admit that this is a very nice solution from the reader's perspective. Note that we will not have documentation for those generated functions, but they are anyway not interesting. Let me see if it works out 👍

@absorbing(MinkowskiSum) = EmptySet

I think you cannot do assignments in macros. So I will introduce @absorbing(MinkowskiSum, EmptySet)

schillic · 2018-02-10T19:39:05Z

Ready from my side.

schillic · 2018-02-11T13:37:48Z

I updated again.

Side note: This macro stuff can be complicated at times. In the end I used a bit of copy-paste because I lost patience.

mforets · 2018-02-11T14:16:54Z

docs/src/lib/utils.md

@@ -9,4 +9,7 @@ end

 ```@docs
 sign_cadlag
+@neutral
+@absorbing
+@neutral_and_absorbing


is @neutral_and_absorbing a remainder?

oops, fixed

mforets

nice work with the macros 👌

schillic · 2018-02-11T15:02:20Z

Thanks for the help and reviews, I think it is much better now.

schillic force-pushed the schillic/162 branch 4 times, most recently from ede59cd to 772504d Compare February 8, 2018 22:08

schillic changed the title ~~[WIP #162] Add macros to automatically create boilerplate code~~ #162 - Add macros to automatically create boilerplate code Feb 8, 2018

schillic added 5 commits February 9, 2018 09:11

add @commutative_neutral macro

f75f6d3

use macro

00df076

add & use @commutative_absorbing macro

5ddb0ca

add & use @commutative_neutral_absorbing macro

8e9f0b8

docs, constructors, restructuring

9348d4a

schillic force-pushed the schillic/162 branch from 3e1fa56 to 9348d4a Compare February 9, 2018 08:50

MinkowskiSum unit tests for neutral/absorbing element

714a8eb

schillic requested a review from mforets February 9, 2018 09:00

mforets reviewed Feb 10, 2018

View reviewed changes

schillic mentioned this pull request Feb 10, 2018

Add examples for constructors and remove code copy #232

Closed

schillic added 3 commits February 10, 2018 19:44

review: do not export, docs

76da433

add neutral/absorbing function and unify macros

345fc24

fix docs

b85d1e4

schillic mentioned this pull request Feb 10, 2018

Add macros to automatically create boilerplate code #162

Closed

mforets reviewed Feb 11, 2018

View reviewed changes

new @neutral and @Absorbing macros

c69313d

schillic force-pushed the schillic/162 branch from ef0e89e to c69313d Compare February 11, 2018 14:20

mforets approved these changes Feb 11, 2018

View reviewed changes

schillic merged commit 7b12631 into master Feb 11, 2018

schillic deleted the schillic/162 branch February 11, 2018 15:01

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

#162 - Add macros to automatically create boilerplate code #228

#162 - Add macros to automatically create boilerplate code #228

schillic commented Feb 8, 2018 •

edited

Loading

schillic commented Feb 8, 2018 •

edited

Loading

mforets Feb 10, 2018

schillic Feb 10, 2018

mforets Feb 10, 2018

schillic Feb 10, 2018

mforets Feb 10, 2018 •

edited

Loading

schillic Feb 10, 2018

mforets Feb 10, 2018 •

edited

Loading

mforets Feb 10, 2018

schillic Feb 10, 2018

mforets Feb 10, 2018 •

edited

Loading

schillic Feb 10, 2018

mforets Feb 10, 2018

schillic Feb 10, 2018

mforets Feb 10, 2018

schillic Feb 10, 2018

mforets Feb 10, 2018

schillic Feb 10, 2018

schillic Feb 10, 2018

mforets Feb 11, 2018 •

edited

Loading

schillic Feb 11, 2018

schillic Feb 11, 2018

schillic commented Feb 10, 2018

schillic commented Feb 11, 2018

mforets Feb 11, 2018

schillic Feb 11, 2018

mforets left a comment

schillic commented Feb 11, 2018


		### Examples

		`@commutative_neutral(ConvexHull, EmptySet)` creates the following functions:


		### Examples

		`@commutative_neutral_absorbing(MinkowskiSum, ZeroSet, EmptySet))` creates the


		- `CartesianProductArray()` -- constructor for an empty Cartesian product
		Constructors:

#162 - Add macros to automatically create boilerplate code #228

#162 - Add macros to automatically create boilerplate code #228

Conversation

schillic commented Feb 8, 2018 • edited Loading

schillic commented Feb 8, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mforets Feb 10, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mforets Feb 10, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mforets Feb 10, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mforets Feb 11, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

schillic commented Feb 10, 2018

schillic commented Feb 11, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mforets left a comment

Choose a reason for hiding this comment

schillic commented Feb 11, 2018

schillic commented Feb 8, 2018 •

edited

Loading

schillic commented Feb 8, 2018 •

edited

Loading

mforets Feb 10, 2018 •

edited

Loading

mforets Feb 10, 2018 •

edited

Loading

mforets Feb 10, 2018 •

edited

Loading

mforets Feb 11, 2018 •

edited

Loading