First attempt at getting origami into Oscar #5345

Sebas777-gif · 2025-09-18T07:11:14Z

We were working for some time on origami functionalities in our fork of Oscar. @fingolfin recommends this draft pull request, although there is probably still work to be done.

This is certainly necessary to make the code calling it type stable

Again necessary to make code using it type stable.

... not as a Vector: a tuple can be allocated on the stack instead of the heap and thus avoid expensive allocations, and in general allows for more optimizations.

Otherwise, cperm creates a fresh parent group each time it is called...

There is (currently, at least) no need to have it mutable, and immutable potentially enables further operations, plus it may catch programming mistakes

- get rid of `degree_list` - make `normal_form` type stable by not reusing `G` as reference to objects of different types - avoid overhead in `cperm` for creating parent groups by creating a single shared parent group at the start. - instead of creating a vector of vector describing permutations; then turning that into a vector of permutations, directly store a vector of permutations (thus reducing the needed allocations) - replace `i^-1 * x * i` by `x^i` which is faster and again avoids allocations (at least when `x` and `i` are permutations, like here)

Improve `staircase_origami` and `normal_form`

…car.jl into origami

origami code tweaks

…car.jl into origami

fingolfin · 2025-09-19T07:33:04Z

experimental/Origami/src/homology_action.jl

@@ -0,0 +1,33 @@
+function homology(o::Origami)
+    return GAP.gap_to_julia(GAP.Globals.HomologyOrigami(GapObj(o)))


All uses of gap_to_julia should be replaced by coercions or type assertions to the desire final type, so that functions are understandable and type stable.

So here, the question would be: what kind of data does HomologyOrigami return, and what would be a good corresponding Julia / Oscar type?

Same for all other calls to gap_to_julia.

This is not a must for a first merge of this PR, but would be nice.

fingolfin · 2025-09-19T07:34:26Z

experimental/Origami/src/normal_origami.jl

+end
+
+function all_normal_origamis_by_degree(n::Int)
+    return GAP.Globals.AllNormalOrigamisByDegree(n)


I am guessing this should really return a Vector{Origami} and the result should be wrapped a such?

E.g. this might work?

Suggested change

return GAP.Globals.AllNormalOrigamisByDegree(n)

return [from_GAP_origami(o) for o in GAP.Globals.AllNormalOrigamisByDegree(n)]

Normal origamis consist of a finite group $G$ and a pair of generators of $G$. See https://ag-weitze-schmithusen.github.io/Origami/doc/chap2.html#X838AD31B7BF71CDF. So I'll have to think a little about how to best handle this in the realm of Oscar.

Also, I think the naming of this function is really confusing; I'll talk to @wsinuds about this.

…tion to return tuples

fingolfin

A ton of little comments here and there. I hope you are not overwhelmed by this... a lot of it can be just applied with a few clicks here (don't forget to pull the changes made on GitHub into your local clone afterwards)

experimental/Origami/src/systoles.jl

experimental/Origami/src/exports.jl

fingolfin · 2025-09-21T23:24:02Z

experimental/Origami/src/generate.jl

+
+# based on Donald E. Knuth The Art of Computer Programming Algorithm H p.300
+# implementation from Sage
+function product_gray_code(m::Vector{Int})::Vector{Tuple{Int, Int}}


Suggested change

function product_gray_code(m::Vector{Int})::Vector{Tuple{Int, Int}}

function product_gray_code(m::Vector{Int})

fingolfin · 2025-09-21T23:25:47Z

experimental/Origami/src/generate.jl

+    end
+
+    for (i, top_seps) in enumerate(cyl_diagram.top)
+        top = []


Type unstable

fingolfin · 2025-09-21T23:26:03Z

experimental/Origami/src/generate.jl

+            lx[j+widths[i]] = j
+        end
+    end
+    lx = perm([x + 1 for x in lx])


Ideally all calls to perm would take the parent symmetric group as first argument

fingolfin · 2025-09-21T23:27:11Z

experimental/Origami/src/generate.jl

+
+    if length(current) == n
+        if sumSoFar <= max
+        push!(combinations, copy(current))  # Add the valid combination to the list


Suggested change

push!(combinations, copy(current)) # Add the valid combination to the list

push!(combinations, copy(current)) # Add the valid combination to the list

fingolfin · 2025-09-21T23:27:34Z

experimental/Origami/src/generate.jl

+        if sumSoFar <= max
+        push!(combinations, copy(current))  # Add the valid combination to the list
+        end
+    return combinations


Suggested change

return combinations

return combinations

Co-authored-by: Max Horn <[email protected]>

…car.jl into origami

fingolfin · 2025-10-06T21:05:46Z

@Sebas777-gif sorry for flooding you with some many comments (note that GitHub doesn't even show them all by default, you need to click on the "show 43 hidden comments" link...).

I think a bunch of them could be just applied with a click from the web interface here. But some probably require a bit of discussion.

Perhaps you (or possibly multiple people from your group) would also be interested in meeting up with us for a day or two to work on this. Happy to chat about that.

Sebas777-gif · 2025-10-08T10:20:51Z

@fingolfin no problem at all, and thanks for your many contributions and suggestions! It is indeed a bit overwhelming though and I've been procrastinating this work for some time now ... I will pitch your idea of meeting up to @wsinuds and @leo-emmerich tomorrow.

Co-authored-by: Max Horn <[email protected]>

lgoettgens

Some more random comments

lgoettgens · 2025-11-11T13:40:55Z

experimental/Origami/src/Origami.jl

+  return (a.h == b.h) && (a.v == b.v)
+end
+
+function Base.hash(o::Origami, h::UInt=0x000000000)


Suggested change

function Base.hash(o::Origami, h::UInt=0x000000000)

function Base.hash(o::Origami, h::UInt)

this should not have a default value, as julia base already provides a one-argument hash method

lgoettgens · 2025-11-11T13:44:21Z

experimental/Origami/src/Origami.jl

+
+function veech_group(O::Origami)
+  # TODO use hashtables or not? Implement modular subgroup?
+  GAP.Globals.ComputeVeechGroupWithHashTables(GapObj(O))


this should have an explicit return. Furthermore, it would probably be good to wrap the return value in an Oscar group type

lgoettgens · 2025-11-11T13:45:40Z

experimental/Origami/src/Origami.jl

+      length(setdiff(degree_list, moved_points(sigma))) +
+      length(setdiff(degree_list, moved_points(sigma * x))) +
+      length(setdiff(degree_list, moved_points(sigma * y)))


Suggested change

length(setdiff(degree_list, moved_points(sigma))) +

length(setdiff(degree_list, moved_points(sigma * x))) +

length(setdiff(degree_list, moved_points(sigma * y)))

number_of_fixed_points(sigma) +

number_of_fixed_points(sigma * x) +

number_of_fixed_points(sigma * y)

lgoettgens · 2025-11-11T13:47:10Z

experimental/Origami/src/Origami.jl

+  cyl_lists = Vector{Vector{Int}}(gap_obj)
+  cyl_tuples = [(l[1], l[2]) for l in cyl_lists]


Suggested change

cyl_lists = Vector{Vector{Int}}(gap_obj)

cyl_tuples = [(l[1], l[2]) for l in cyl_lists]

cyl_tuples = Vector{Tuple{Int, Int}}(gap_obj)

this could work as well (I haven't tested it)

lgoettgens · 2025-11-11T13:49:07Z

experimental/Origami/src/generate.jl

+    error("Stratum not supported!")
+  end
+
+  file_path = joinpath(dirname(@__FILE__), "..", "cylinder_diagrams", file_name)


Suggested change

file_path = joinpath(dirname(@__FILE__), "..", "cylinder_diagrams", file_name)

file_path = joinpath(@__DIR__, "..", "cylinder_diagrams", file_name)

lgoettgens

Loading the two GAP packages produces a lot of warnings and errors:

#E  component `License' must be bound to a nonempty string containing an SPDX ID
#E Validation of package modulargroup from /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/ModularGroup-2.0.0/\
 failed
#E  component `License' must be bound to a nonempty string containing an SPDX ID
#E Validation of package origami from /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/ failed
#I  DeclareGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/cyclic-torus-covers.gd:67
#I  DeclareGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/cyclic-torus-covers.gd:72
#I  DeclareGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/cyclic-torus-covers.gd:81
#I  DeclareGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/cyclic-torus-covers.gd:100
#I  DeclareGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/cyclic-torus-covers.gd:110
#I  DeclareGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/cyclic-torus-covers.gd:123
#I  DeclareGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/cyclic-torus-covers.gd:136
#I  DeclareGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/cyclic-torus-covers.gd:151
#I  The package 'ArangoDBInterface' is currently not installed. Without this package, the origami database is not available.
#I  InstallGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/origami.gi:346
#I  InstallGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/origami.gi:366
#I  InstallGlobalFunction: too many arguments in /home/lgoe/.julia/scratchspaces/c863536a-3901-11e9-33e7-d5cd0df7b904/gap_packagedir_v4.14/pkg/Origami-2.0.1/lib/origami.gi:371
#I  The package 'ArangoDBInterface' is currently not installed. Without this package, the origami database is not available.

I think most (or even all) of them are already fixed in the packages thanks to @fingolfin, but they require a release of the packages to end up here. This needs to be addressed before this PR can be merged.

fingolfin · 2025-11-11T15:42:42Z

@lgoettgens we are well aware of these issues. @Sebas777-gif and others from Saarbrücken are going to visit end of the month in Kaiserslautern and we'll try to resolve most / all of this.

Pascalonyy and others added 30 commits November 6, 2023 11:32

Start of Origami

b14aba0

Origami Object

b4e082e

finished origami struct, added basic functionality

fea5c58

everything up to OrigamiEquivalent

3ab88a4

implemented some function in pure oscar

a64bb20

automorphisms

3c2b52f

Merge branch 'oscar-system:master' into origami

9e44030

implemented deck_group

f8f6a3e

implemented sl2z action, and normal form for origamis

3b03190

some coding style fixes and special origamis

b8d3d8a

Make cycle_length type stable

4419eeb

This is certainly necessary to make the code calling it type stable

Ensure struct Origami contains concrete types

f07b459

Again necessary to make code using it type stable.

Store minimal_cycle_lengths as tuple

1755e79

... not as a Vector: a tuple can be allocated on the stack instead of the heap and thus avoid expensive allocations, and in general allows for more optimizations.

Reuse a single parent in cperm calls

a397ffe

Otherwise, cperm creates a fresh parent group each time it is called...

Make struct Origami immutable

8ec034e

There is (currently, at least) no need to have it mutable, and immutable potentially enables further operations, plus it may catch programming mistakes

Add GAP.GapObj HACK (temporary, should be removed eventually)

41f39c7

Optimize cycle_length

60bba04

Avoid some more allocations

81c8ec0

Merge pull request #1 from fingolfin/mh/origami

bae15ad

Improve `staircase_origami` and `normal_form`

generating origamis

b1adc7f

Merge remote-tracking branch 'upstream/master' into origami

2d4a994

finished special origamis

35d110f

working on homology action

bbf0952

homologyaction

2971c2b

added some more functionality

885cbb0

normal origamis

ee975fd

fixed types

2ae6f1d

Merge branch 'origami' of https://github.com/AG-Weitze-Schmithusen/Os…

643e0be

…car.jl into origami

Merge branch 'oscar-system:master' into origami

14e499a

Sebas777-gif and others added 3 commits September 19, 2025 09:14

Merge pull request #5 from fingolfin/mh/origami-updated-and-tweaks

f5f81a4

origami code tweaks

Started documentation

6bbe64a

Merge branch 'origami' of https://github.com/AG-Weitze-Schmithusen/Os…

00f7518

…car.jl into origami

fingolfin reviewed Sep 19, 2025

View reviewed changes

Sebas777-gif added 2 commits September 19, 2025 09:40

Fixed some bugs that were just introduced; sorry for this oversight

3a21f5f

Fixed exports; removed unnecessary wrapper; change automorphisms func…

bfe32b8

…tion to return tuples

fingolfin reviewed Sep 21, 2025

View reviewed changes

Sebas777-gif and others added 3 commits September 25, 2025 18:56

Better array initialization in systolic_ratio_bigger_one_over_pi_in_h11

58dbbeb

Co-authored-by: Max Horn <[email protected]>

Added missing newline at end of exports.jl

ba23423

Merge branch 'origami' of https://github.com/AG-Weitze-Schmithusen/Os…

a4c1b6c

…car.jl into origami

Sebas777-gif and others added 13 commits November 5, 2025 18:06

replace gap_to_julia call with Float64 conversion

59cacfe

Remove unnecessary "export GapObj"

d147bfa

Co-authored-by: Max Horn <[email protected]>

Remove unnecessary special case optimization for GapObj

d160717

Co-authored-by: Max Horn <[email protected]>

Cleaner write-up of the genus formula

8d497ab

Co-authored-by: Max Horn <[email protected]>

Use inplace sort to avoid allocating

cc71cd4

Co-authored-by: Max Horn <[email protected]>

Fix formatting

f874491

Fix some more formatting

2841ec7

More efficient computation of translations via stabilizer

e7794e9

Remove clarified TODO

8e3f358

Nicer code for is_hyperelliptic

91d6129

Made normalform_conjugators type stable

ec712be

Nicer syntax in normalform_conjugators

e3f5643

More elegant code for point_reflections

ca41c8e

lgoettgens reviewed Nov 11, 2025

View reviewed changes

lgoettgens requested changes Nov 11, 2025

View reviewed changes

lgoettgens mentioned this pull request Nov 11, 2025

Load gap packages from artifacts AG-Weitze-Schmithusen/Oscar.jl#6

Open

		@@ -0,0 +1,33 @@
		function homology(o::Origami)
		return GAP.gap_to_julia(GAP.Globals.HomologyOrigami(GapObj(o)))

	return GAP.Globals.AllNormalOrigamisByDegree(n)
	return [from_GAP_origami(o) for o in GAP.Globals.AllNormalOrigamisByDegree(n)]

	function product_gray_code(m::Vector{Int})::Vector{Tuple{Int, Int}}
	function product_gray_code(m::Vector{Int})

	push!(combinations, copy(current)) # Add the valid combination to the list
	push!(combinations, copy(current)) # Add the valid combination to the list

	function Base.hash(o::Origami, h::UInt=0x000000000)
	function Base.hash(o::Origami, h::UInt)

		cyl_lists = Vector{Vector{Int}}(gap_obj)
		cyl_tuples = [(l[1], l[2]) for l in cyl_lists]

	cyl_lists = Vector{Vector{Int}}(gap_obj)
	cyl_tuples = [(l[1], l[2]) for l in cyl_lists]
	cyl_tuples = Vector{Tuple{Int, Int}}(gap_obj)

	file_path = joinpath(dirname(@__FILE__), "..", "cylinder_diagrams", file_name)
	file_path = joinpath(@__DIR__, "..", "cylinder_diagrams", file_name)

First attempt at getting origami into Oscar #5345

Are you sure you want to change the base?

First attempt at getting origami into Oscar #5345

Uh oh!

Conversation

Sebas777-gif commented Sep 18, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fingolfin left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

fingolfin commented Oct 6, 2025

Uh oh!

Sebas777-gif commented Oct 8, 2025

Uh oh!

lgoettgens left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lgoettgens left a comment

Choose a reason for hiding this comment

Uh oh!

fingolfin commented Nov 11, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants