Use Arachne to vary line width #1210

Ghostkeeper · 2020-03-04T08:28:07Z

For now, this is just a WIP comparison to be able to offer review comments on libArachne.

Do not merge!

4.8 behaviour	Arachne behaviour

See CURA-7235.

src/BeadingOrderOptimizer.h

src/OutlineAccuracyBeadingStrategy.cpp

src/OutlineAccuracyBeadingStrategy.h

src/SingleBeadBeadingStrategy.cpp

src/TestGeometry/Spiky.h

src/utils/VoronoiUtils.cpp

src/voronoi_visual_utils.hpp

Ghostkeeper · 2020-03-10T08:21:31Z

Can we move these files into a separate folder? There are a lot of beading strategy classes in the main source folder now.

Ghostkeeper

I didn't get nearly as far as I hoped. It's tough to review with so little documentation. So I'll continue tomorrow.

At the moment this branch doesn't compile due to a billion missing imports.

src/ToolpathWriter.cpp

src/ToolpathWriter.h

src/Statistics.h

src/TestGeometry/Jin.h

src/TestGeometry/Pika.h

src/SkeletalTrapezoidation.cpp

Ghostkeeper · 2020-03-11T16:33:22Z

src/SkeletalTrapezoidation.cpp

+            continue;
+        }
+
+        // copy start to end edge to graph


? Sentence? Are we copying twice, or...?

Test reply! If github works properly you should see his!

Ghostkeeper · 2020-03-11T16:33:58Z

src/SkeletalTrapezoidation.cpp

+        node_t* starting_node = vd_node_to_he_node[starting_vd_edge->vertex0()];
+        starting_node->data.distance_to_boundary = 0;
+        // starting_edge->prev = nullptr;
+//         starting_edge->from->data.distance_to_boundary = 0; // TODO


Oops, the library is not finished yet!

src/SkeletalTrapezoidation.h

src/SkeletalTrapezoidation.cpp

Ghostkeeper · 2020-03-13T08:46:24Z

src/SkeletalTrapezoidation.cpp

+        edge_locator.emplace(&*edge_it, edge_it);
+    for (auto node_it = graph.nodes.begin(); node_it != graph.nodes.end(); ++node_it )
+        node_locator.emplace(&*node_it, node_it );
+    auto safelyRemoveEdge = [this, &edge_locator](edge_t* to_be_removed, std::list<edge_t>::iterator& current_edge_it, bool& edge_it_is_updated)


Consider making this a normal function of the HalfEdgeGraph.

Like i said for the other comment, this one works on the edge_locator collection defined in the function rather than the graph itself.

Moving the edge_locator to the edge class is a bit overkill imho, since it only occurs within this function (the surrounding one, not the closure).

src/SkeletalTrapezoidation.cpp

Ghostkeeper · 2020-03-13T08:49:29Z

src/SkeletalTrapezoidation.cpp

+
+    for (auto edge_it = graph.edges.begin(); edge_it != graph.edges.end();)
+    {
+        if (edge_it->prev)


Document: This is not the start of a quad then, right?

Ghostkeeper · 2020-03-13T08:52:17Z

src/SkeletalTrapezoidation.cpp

+        bool quad_mid_is_removed = false;
+        if (quad_mid && should_collapse(quad_mid->from, quad_mid->to))
+        {
+            assert(quad_mid->twin);


I mean... This will happen by itself on line 647 😛

src/SkeletalTrapezoidation.cpp

Ghostkeeper · 2020-03-17T10:24:33Z

src/utils/polygon.cpp

+            Point c = poly[(i + 1) % poly.size()];
+            if (LinearAlg2D::pointIsLeftOfLine(b, a, c) >= 0)
+            {
+                poly.remove(i);


We might want to copy the data to a separate polygon and then std::swap for better theoretical performance and simplification of this code.

I've implemented this now.

src/utils/polygonUtils.cpp

src/utils/polygonUtils.h

src/utils/polygonUtils.cpp

src/voronoi_visualizer.cpp

Ghostkeeper · 2020-03-17T10:39:41Z

So here are the broad strokes of things that still need to happen for this feature before we can start testing:

Move the beading strategies (and perhaps all of Arachne) to a separate folder.
Convert the output to discrete segments (though there is code to do this, it's not in the form that we need it).
Fill in the missing documentation and review again with more documentation where necessary.
FIx code style.
Remove the debug code.
Remove the beading strategies that are not useful for 3D printing (but were just used for comparison to other techniques).
Process all of these specific review comments for individual improvements we can make to the code.
Instead of generating outlines normally, call upon Arachne to do this for us.
Remove the gap filling technique.

src/SkeletalTrapezoidation.cpp

src/SkeletalTrapezoidation.h

rburema · 2020-05-12T21:19:21Z

src/utils/HalfEdgeNode.h

@@ -48,6 +48,21 @@ class HalfEdgeNode
        }
        return odd_path_count > 2;
    }
+
+    bool isMarked() const


Proposal: Merge isMultiIntersection and isMarked into a single function that checks whether the isMarked count of going around an edge is larger-or-equal than '2 or 1, respectively', which can be the argument.

(We could also muck around with writing a sort of map-around-the-vertex function that takes a closure as argument, but that's probably overkill.) (Anyway, at least isMultiIntersection can also have an early-out and then return false just like isMarked.)

CMakeLists.txt

src/BeadingStrategy/SingleBeadBeadingStrategy.cpp

Ghostkeeper

These are the comments that have not yet been resolved but were marked as "outdated" by Github due to code being changed or moved around it, except some more documentation comments that I didn't bother to mention.

src/SkeletalTrapezoidation.h

Ghostkeeper · 2020-05-19T15:56:00Z

src/SkeletalTrapezoidation.h

+     * \param going_up Whether we are traveling in the upward direction as seen from the \p origin_transition. If this doesn't align with the direction according to the R diff on a consecutive edge we know there was a local optimum
+     * \return whether the origin transition should be dissolved
+     */
+    std::list<TransitionMidRef> dissolveNearbyTransitions(edge_t* edge_to_start, TransitionMiddle& origin_transition, coord_t traveled_dist, coord_t max_dist, bool going_up, std::unordered_map<edge_t*, std::list<TransitionMiddle>>& edge_to_transitions);


Please use std::vector rather than std::list, at least for the return type. Vector is going to give us much more performance, and these transitions are going to be quite common. See my previous comment here: #1210 (comment)

... except that he does a splice on at least one of the results, which is available for lists only (would be inefficient for vectors).

to_be_dissolved.splice(to_be_dissolved.end(), to_be_dissolved_here); // Transfer to_be_dissolved_here into to_be_dissolved

Ghostkeeper · 2020-05-19T16:04:55Z

src/SkeletalTrapezoidationGraph.cpp

+
+void SkeletalTrapezoidationGraph::makeRib(edge_t*& prev_edge, Point start_source_point, Point end_source_point, bool is_next_to_start_or_end)
+{
+    Point p = LinearAlg2D::getClosestOnLineSegment(prev_edge->to->p, start_source_point, end_source_point);


If I understood correctly, we can use getClosestOnLine here, rather than getClosestOnLineSegment. We've already checked bounds of the projection so we don't need to do that again. See my previous comment here: #1210 (comment)

In the sense that it wouldn't make much sense to have a rib that is not connected to the line segment (partially) defining it, and we have already committed to making the rib here, right?

src/utils/VoronoiUtils.cpp

src/SkeletalTrapezoidationGraph.cpp

src/SkeletalTrapezoidation.cpp

Sophist-UK · 2020-08-31T23:19:56Z

Is it possible to have a description and some diagrams to show how this PR is intended to work, so that we can comment on the Pros and Cons of the intended solution - which might help to get to a great solution quicker or might be simply an annoying distraction?

rburema · 2020-09-01T09:24:32Z

@Sophist-UK The closest thing to an explanation right now is probably this paper which also includes figures and diagrams -- the initial work and lions' share for this PR was done by the first author.

[CURA-8354](redux) unit-tests for 'Fix width adjust simplify'

This reverts commit 8695be6. That commit was making the wrong assumption that holes are not visible from the outside. They are. And there might be other shapes inside of the hole, such as nested rings. Internal holes can have any shape, and don't necessarily need to be convex. They can have protruding parts that need to be avoided. We need to treat inside walls just the same way as outside walls. Contributes to issue CURA-8117.

Conflicts: src/skin.cpp -> Code changed in an area that was removed. src/utils/polygon.h -> Include statements changed while we added one for Arachne nearby.

This is to fix the case, where: - We are travelling from one part to another. - The start (or end) part is not convex. - The start (or end) part is so thin that traversing it through the optimal combing boundary (far from the walls) is not possible, but traversing through the minimum combing boundary (slightly wider, but close to the outer wall) is possible. In this case, what happened is that it was trying to find the spot inside the starting part closest to the destination and would try to comb there. But because the part is so thin, it failed to find a comb path with the optimal combing boundary. Instead, it would just not add a combing path. It would then move outside on the starting point which is nearby a hole in the model. The hole is closed, so there is no travel path through air to the destination part, so that will just take a straight line to the destination part. This straight line intersects multiple times with the starting part, which could have been avoided. This solution causes the combing to try the minimum combing boundary if the optimal combing boundary fails. Previously this was only used for the case where we're travelling within the same part. Now it'll also be used to prepare travelling to different parts. The CPU time needed for this is only bigger if this case actually occurs. For normal, sufficiently thick parts there should be no increase in slicing time. Contributes to issue CURA-8117.

Rather than the minimum ones that were used to calculate the crossing with. Contributes to issue CURA-8117.

Forgot to update this one in the case of the end part. Contributes to issue CURA-8117.

…imum_bound If combing inside fails, allow combing through minimum bound

Turns out that we could still use the majority of the spreadDotsAera that T. wrote for Lightening fill. My bad! Co-authored-by: BagelOrb <t.kuipers@ultimaker.com>

Conflicts: src/FffPolygonGenerator.cpp -> Maximum layer count fix src/InsetOrderOptimizer.cpp -> Fix for seam position which is no longer located here. src/pathPlanning/Comb.cpp -> Several combing fixes discarded because combing is pretty much revamped in Arachne. src/skin.cpp -> Gap filling area fix, but gap filling is removed in Arachne. src/utils/PolygonProximityLinker.cpp -> Wall overlap compensation, but removed in Arachne. src/utils/algorithm.h -> Headers rearranged. src/utils/math.h -> Addition of some helper functions changed headers. src/utils/polygonUtils.h -> Code style

As discussed during the eCCB 12/11/21 this factor makes more sense as a Ratio.

…e_rebased

Instead of the minimum. So previously a value of 2 would make it spread the error over 2 walls for an even number, or 3 for an odd. Now it makes it spread the error over 2 walls for an even number or 1 for an odd. This means that the minimum allowed value becomes 2, because for an even wall count it needs to place the error somewhere. Done as a 5 minute fix.

Ligning infill was first generated, then applied to the part. However, the enitre fill would be applied to each part. If a model consisted of more than one part, the lightning paths would be retraced for each part of the model in that layer. Now it limits the output to the outlines that are currently processed. It's relatively efficient, since it only has to check one node of each tree against the outlines of a single part. fixes Cura/#10793

Now the length is measured along the branches, which is also good. Co-authored-by: Ghostkeeper <Ghostkeeper@users.noreply.github.com>

part of CURA-8702

Use functions that already exist instead of implementing new ones that mess up. part of CURA-8702

This reverts commit 55d0e58. That commit was done as an unsure 'well if it can't harm' kind of way in order to _maybe_ halfway fix a bug that wasn't possibly wasn't affecting our prints and now appears to have been fixed in other ways, as I can't even reproduce it anymore. However, the now reverted fix may have casued the issue in CURA-8713 -- retracts in gcode preventing the UM2+ from finishing prints.

Ghostkeeper · 2021-12-14T14:06:13Z

Is not often that one gets to merge 1377 commits in one push!

nallath requested changes Mar 6, 2020

View reviewed changes

Ghostkeeper commented Mar 11, 2020

View reviewed changes

Ghostkeeper commented Mar 17, 2020

View reviewed changes

Ghostkeeper mentioned this pull request Apr 7, 2020

Stuttering in curved movements due to Wall Overlap Compensation Ultimaker/Cura#6676

Closed

rburema marked this pull request as draft April 10, 2020 11:21

rburema mentioned this pull request Apr 10, 2020

Compiles & UnitTest Fix on Win for libAracnhe branch #1236

Merged

Ghostkeeper mentioned this pull request Apr 28, 2020

unique Benchy defect with Cura 4.6 beta or 4.6 final Ultimaker/Cura#7611

Closed

Ghostkeeper mentioned this pull request May 7, 2020

Duplicate inside wall line when printing a wall of thickness 3x line width Ultimaker/Cura#7656

Closed

nallath force-pushed the libArachne_rebased branch from b5c9fed to fe8551f Compare May 7, 2020 13:12

Ghostkeeper mentioned this pull request May 7, 2020

[4.6.1] [4.5] nozzle going over random line segments after printing them, 'ironing' them despite ironing disabled and it being midway up a print (not an upper layer) Ultimaker/Cura#7687

Closed

rburema requested changes May 12, 2020

View reviewed changes

Ghostkeeper commented May 19, 2020

View reviewed changes

Ghostkeeper mentioned this pull request May 26, 2020

Overextrusion on thin line ceiling. Ultimaker/Cura#7818

Closed

rburema reviewed May 27, 2020

View reviewed changes

src/SkeletalTrapezoidation.cpp Outdated Show resolved Hide resolved

rburema reviewed May 27, 2020

View reviewed changes

src/SkeletalTrapezoidation.cpp Outdated Show resolved Hide resolved

This was referenced Jun 10, 2020

Selective Horizontal Expansion Ultimaker/Cura#7882

Closed

Inconsistent surfaces caused by WallOverlapComputation being used in random directions #1266

Closed

rburema mentioned this pull request Jul 15, 2020

Implement variable line width. Ultimaker/Cura#8065

Closed

Ghostkeeper mentioned this pull request Jul 16, 2020

Variable Line Width Ultimaker/Cura#5933

Closed

Ghostkeeper mentioned this pull request Aug 4, 2020

3.6.0 - "Fill Gaps Between Walls" is causing printing problems Ultimaker/Cura#5110

Closed

Ghostkeeper mentioned this pull request Aug 14, 2020

Wall Collapse #1187

Closed

rburema mentioned this pull request Sep 17, 2020

Improvement to gap filling in the works? Ultimaker/Cura#8396

Closed

Ghostkeeper mentioned this pull request Sep 21, 2020

Math error positioning first layer lines. Ultimaker/Cura#8404

Closed

nallath and others added 22 commits September 15, 2021 16:02

Merge pull request #1501 from Ultimaker/fix_width_adjust_simplify

781a7b0

[CURA-8354](redux) unit-tests for 'Fix width adjust simplify'

Merge branch 'master' into libArachne_rebased

ca64bdb

Conflicts: src/skin.cpp -> Code changed in an area that was removed. src/utils/polygon.h -> Include statements changed while we added one for Arachne nearby.

Actually use optimum combing bounds for optimum attempt

aec4380

Rather than the minimum ones that were used to calculate the crossing with. Contributes to issue CURA-8117.

Update succeeded flag if combing succeeds in the end

b66d8a1

Forgot to update this one in the case of the end part. Contributes to issue CURA-8117.

Merge pull request #1503 from Ultimaker/CURA-8117_combing_through_min…

dfbcaf9

…imum_bound If combing inside fails, allow combing through minimum bound

Merge branch 'libArachne_rebased' into merge_arachne_20211015

d208d6d

Partially revert merge.

549447e

Turns out that we could still use the majority of the spreadDotsAera that T. wrote for Lightening fill. My bad! Co-authored-by: BagelOrb <t.kuipers@ultimaker.com>

Merge branch 'merge_arachne_20211015' into libArachne_rebased

409a30b

back pressure compensation as Ratio

dc95131

As discussed during the eCCB 12/11/21 this factor makes more sense as a Ratio.

Merge branch 'flow_equalization_width_factor_as_ratio' into libArachn…

e1ebc08

…e_rebased

Fix; Measure from local root location instead of prev. opt.

8bec65d

Now the length is measured along the branches, which is also good. Co-authored-by: Ghostkeeper <Ghostkeeper@users.noreply.github.com>

Update documentation.

0a56101

part of CURA-8702

Simpler way of finding tree-point insude part.

23515d0

Use functions that already exist instead of implementing new ones that mess up. part of CURA-8702

Fix: Compiler couldn't find M_PI_4 for some reason.

c95bcd7

Merge tag '4.12.1' into libArachne_rebased

f379847

Fix merge, use slightly different outline instead.

5cd30a2

Ghostkeeper merged commit df96b2f into master Dec 14, 2021

This was referenced Dec 30, 2021

Flow rate exceeds theoretical maximum about 2 times Ultimaker/Cura#11093

Closed

Fix polyline handling #1362

Closed

Huge gaps in the center of concentric bottom layers Ultimaker/Cura#11128

Closed

This was referenced Feb 1, 2022

Optimize Filling Gaps Between Walls Ultimaker/Cura#8054

Closed

Cura 4.13.0 Thin wall issue Ultimaker/Cura#11320

Closed

casperlamboo mentioned this pull request Apr 10, 2024

Missing lines causing others to be printed mid-air Ultimaker/Cura#18397

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Use Arachne to vary line width #1210

Use Arachne to vary line width #1210

Ghostkeeper commented Mar 4, 2020 •

edited by jellespijker

Loading

Ghostkeeper commented Mar 10, 2020

Ghostkeeper left a comment

Ghostkeeper Mar 11, 2020

rburema May 28, 2020

Ghostkeeper Mar 11, 2020

Ghostkeeper Mar 13, 2020

rburema May 27, 2020

Ghostkeeper Mar 13, 2020

Ghostkeeper Mar 13, 2020

Ghostkeeper Mar 17, 2020

Ghostkeeper Jun 15, 2020

Ghostkeeper commented Mar 17, 2020

rburema May 12, 2020

Ghostkeeper left a comment

Ghostkeeper May 19, 2020

rburema May 27, 2020

Ghostkeeper May 19, 2020

rburema May 27, 2020

Sophist-UK commented Aug 31, 2020

rburema commented Sep 1, 2020

Ghostkeeper commented Dec 14, 2021

Use Arachne to vary line width #1210

Use Arachne to vary line width #1210

Conversation

Ghostkeeper commented Mar 4, 2020 • edited by jellespijker Loading

Ghostkeeper commented Mar 10, 2020

Ghostkeeper left a 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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Ghostkeeper commented Mar 17, 2020

Choose a reason for hiding this comment

Ghostkeeper left a 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

Sophist-UK commented Aug 31, 2020

rburema commented Sep 1, 2020

Ghostkeeper commented Dec 14, 2021

Ghostkeeper commented Mar 4, 2020 •

edited by jellespijker

Loading