Multiple track dragging development

[Stuart Tyler]

Hi,

I have been busy extending pcbnew to include multiple track dragging functionality. I have made some reasonable progress in getting to understanding the problem and learning new stuff in c++ since I last used it, and lots of KiCad internals, and fun geometry.

Anyway my code is a mess, but my first objective was to understand the geometry, which I have partly sorted out, and there are many corner cases still to resolve. My major issue is that the method EDIT_TOOL::doEndMoveTracksSelection() that does all the work needs to make/break some track and push these back to the pcb. 

I have been through the code several times and a bit confused on how to do this at all. I have looked at other pcb tools and get mixed up in various methods at the low level LINE_PLACER, and then how to get this into the pcb through BOARD.

So in my fork of KiCad (https://gitlab.com/stuarttyler/kicad), branch MultiTrackEndMove, I have modified edit_tool_move_fct.cpp and lines 1244/1312/1332 the existing track is stolen to see an output. This should not happen and new tracks should be formed.

EDIT_TOOL::doEndMoveTracksSelection() started out as a copy of EDIT_TOOL::doMoveSelection()

This was good as it showed me how to handle multiple selections. But it does not do any track additions, so I cannot seem to work that part out.

Is this the best way to share my request? By posting a link to my fork and point to my code?

Currently I have added a new menu item and key press ctrl-M. When there are multiple traces selected, the ends closest to the cursor all move in the direction commanded. There are a few bugs still to resolve, but I think Im about 60% there with the major 3 situations identified. There are a number of fold back situations that need sorting out.

First of all I need help on adding/breaking a trace inside EDIT_TOOL::doEndMoveTracksSelection()

In summary: 

sel_items is a vector of BOARD_ITEM*

    std::vector<BOARD_ITEM*> sel_items;

item get picked out using a cpp for loop to EDA_ITEM*

                for( EDA_ITEM* item : sel_items )

I have worked out how to modify the PCB_TRACK start and end locations

 dynamic_cast<PCB_TRACK*>( item )->SetEnd( myNewEndLocation );

But I cannot work out how to break this track into 2, or add a new track to the board.

What I think I want to do is to make a new BOARD_ITEM or EDA_ITEM but I am obviously missing something on PCB_TRACK manipulation and EDIT_TOOL.

Please help.

Thanks Stuart

PS - I know my code is very messy and horrible. I am currently just getting to terms with all the new geometry and method/functions/internals.

PSS - I plan on having something more sensible/presentable in a few weeks.

[Stuart Tyler]

I have updated the code and I think the geometry manipulation is now close.

Please note the move multiple tracks key press that I have coded up is Alt-M, not Ctrl-M.

 

Just need some help on adding tracks to a pcb so that the solution can be fully visualised. There are probably a few other corner cases that need handling yet, but its probably useable once the track adds are fixed.

Ta

Stuart

[Stuart Tyler]

Hi, I have tried to add a track, but its not appearing on the board.

Here is my code in my code that does the track add. Is in edit_tool_move_fct.cpp (1259)

                        printf( "*** Adding new track to board ***\r\n" );
                        std::vector<BOARD_ITEM*> bulkAddedItems;
                        PCB_TRACK*  pcbTrackOfItem = dynamic_cast<PCB_TRACK*>( item );
                        auto        newTrack = new PCB_TRACK( pcbTrackOfItem );
                        BOARD_ITEM* newBoardItem = dynamic_cast<BOARD_ITEM*>( newTrack );
                        newBoardItem->SetPosition( nearestEndLocation[itemCount] );
                        dynamic_cast<PCB_TRACK*>( newBoardItem )->SetEnd( furtherestEndLocation[itemCount] );
                        board->Add( newBoardItem, ADD_MODE::BULK_INSERT, false );
                        bulkAddedItems.push_back( newBoardItem );
                        board->FinalizeBulkAdd( bulkAddedItems );
.. and later

                        newBoardItem->SetModified();

Can anyone spot the issue with my code and why I would not get a track showing up?

Ta

Stuart

[Jon Evans]

Hi Stuart,

Some general comments on this project:

1) This functionality should be built into the router, not into the edit routines outside the router IMO.  So this should be implemented as part of DRAGGER probably.  If you look at how the rest of the router works, you should be able to re-use all its functionality for creating and modifying items in the board.

2) The functionality you're looking for is at some level similar to the work that is done by the PNS component dragger (since it has to be able to reroute multiple different tracks), so you might get some additional hints from that class.

3) There is no need for a new action / hotkey.  The existing actions should be able to handle single-track vs multi-track dragging based on what the selection contains.

Best,

Jon

--
You received this message because you are subscribed to the Google Groups "KiCad Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to devlist+u...@kicad.org.
To view this discussion on the web visit https://groups.google.com/a/kicad.org/d/msgid/devlist/64d0d1d5-6a24-41c4-8231-a4a721f267c7n%40kicad.org.

[Stuart Tyler]

Hi Jon,

Thanks for the feedback. Yes, I agree with all your points. I started off copying drag track but backed out of this as I got stuck accessing the multiple selections, but totally agree with your points.

I have not looked through the component dragger as yet and will visit that over the next couple of days. I think your comparison is correct - I will attempt to follow its track movement and addition style - I now understand the algorithm and what is needed at least.

Appreciate the pointer to looking at component dragger.

Regards

Stuart Tyler

[Tomasz Wlostowski]

On 18/07/2023 20:06, Stuart Tyler wrote:
> Hi Jon,
>
> Thanks for the feedback. Yes, I agree with all your points. I started
> off copying drag track but backed out of this as I got stuck accessing
> the multiple selections, but totally agree with your points.
>
> I have not looked through the component dragger as yet and will visit
> that over the next couple of days. I think your comparison is correct -
> I will attempt to follow its track movement and addition style - I now
> understand the algorithm and what is needed at least.
>
> Appreciate the pointer to looking at component dragger.
>

Hi Stuart,

I encourage you to have a look at this branch:
https://gitlab.com/twlostow/kicad/-/commits/tom-multidrag-demo-clean

It contains a very shitty multi-drag algorithm - but my intention there
was not to develop one, but to show how I would see it integrated in the
router & tool stack of pcbnew. I hope you can build a decent multi-trace
dragger on top of this code.

I'll be happy to answer any questions you might have, but be prepared
for some delay with responses.

Regards,
Tom

>> <https://gitlab.com/stuarttyler/kicad>), branch

>> <mailto:devlist+u...@kicad.org>.

>> To view this discussion on the web visit

>> https://groups.google.com/a/kicad.org/d/msgid/devlist/64d0d1d5-6a24-41c4-8231-a4a721f267c7n%40kicad.org <https://groups.google.com/a/kicad.org/d/msgid/devlist/64d0d1d5-6a24-41c4-8231-a4a721f267c7n%40kicad.org?utm_medium=email&utm_source=footer>.

>
> --
> You received this message because you are subscribed to the Google
> Groups "KiCad Developers" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to devlist+u...@kicad.org

> <mailto:devlist+u...@kicad.org>.

> To view this discussion on the web visit

> https://groups.google.com/a/kicad.org/d/msgid/devlist/4b954b95c8997.e6c27fd20fe7b%40gmail.com <https://groups.google.com/a/kicad.org/d/msgid/devlist/4b954b95c8997.e6c27fd20fe7b%40gmail.com?utm_medium=email&utm_source=footer>.

[Stuart Tyler]

Hi Tom,

Was a great catch up with the developers in Spain.

Thanks for prompting me about the below email and branch. I will get this and look at it during this week and attempt to add the algorithm there.

Talk soon.

Regards

Stuart

Get BlueMail for Android

[Stuart Tyler]

Ok, I have spent some time looking at the branch, getting it to compile and understanding the multi-drag extension code. As you say this is not an implementation, but a demo of the integration.

I have an algorithm implementation that I want to try out.

What I think I next need to do is to port my code into the branch that you have provided. Are you expecting me to commit my edits into your Gitlab repo? I am happy to do that, just checking you want me to do that.

My main issue is my poor C++ reading and coding skills. My methods are very hacky and I think I have succeeded in proving the algorithm is correct, but I am not comfortable in implementing it correctly. I am happy for others proficient in C++ to correct my code. 

What I spend much time on is 'How do I ....' issues - like, how do I go through the list of currently selected track and find the closest one to the aP? I know what I want to do to the selected list of tracks, just how to code that in C++ is tricky. What I would like to know is: Is there a C++ coder in the team that would like to spend some time assisting me? I am currently in Europe and happy to do a few evenings on knocking this over?

If there is, I am happy to work on this as a team. I use Slack or MSTeams or any other platform that would assist in implementing this together. If there is no-one available then I will continue to work away slowly on this. 

My repo with edit_tools_move_fct.cpp has the core algorithm in it. I am going to head off and see if I can transport this code into Toms Github example code, and do local commits until I am told I can commit to Toms GitHub. Hopefully someone with decent C++ skills can help me out and assist in closing this out.

I understand that 8.00 release is near, so I will continue working local until I hear something from the team, and I understand if this is not soon. I will shout out if I work it out on my own.

Regards

Stuart Tyler

[Stuart Tyler]

My first sticking point on understanding how to transport my code:

I originally interfaced to the selected parts through selection which was loopable as an EDA_ITEM and could dynamic_cast to PCB_TRACK and access the track detail.

*** start of code snip ***

    for( EDA_ITEM* item : selection )

    {

        ++count;

        BOARD_ITEM* boardItem = dynamic_cast<BOARD_ITEM*>( item );

        FOOTPRINT*  footprint = dynamic_cast<FOOTPRINT*>( item );

        PCB_TRACK*  trackItem = dynamic_cast<PCB_TRACK*>( item );

        if (trackItem)

        {

            printf( "Item %d was able to be cast to a track start %d , %d end %d , %d.\r\n", count,

*** end of code snip ***

But are now passed in as class ITEM. 

*** start of new code snip ***

    for( ITEM* item : aPrimitives.Items() )

    {

*** end of new code snip ***

I am not too sure on how to get to the PCB_TRACK class when given the ITEM class.

Getting to the PCB_TRACK class is important to access all the track end point data for the algorithm computations. Help appreciated.

Regards

Stuart Tyler

[Jon Evans]

Hi Stuart,

The PNS router uses a proxy model of the board, and ITEM is the base class for these proxy representations.  See pns_item.h

You can get back to the BOARD_ITEM using Parent(), but you may not need to; the geometry of tracks will be represented inside the proxy ITEMs.=

-Jon

[Stuart Tyler]

Thanks Jon,

Yes, that worked. My original code math is now working. Thanks.

    for( ITEM* item : aPrimitives.Items() )

    {

        ++count;

        BOARD_ITEM* boardItem = dynamic_cast<BOARD_ITEM*>( item->Parent() );

 //       FOOTPRINT*  footprint = dynamic_cast<FOOTPRINT*>( item->Parent() );

        PCB_TRACK*  trackItem = dynamic_cast<PCB_TRACK*>( item->Parent() );

Regards

Stuart Tyler