[Question] Is there a way/an option to "force" position coordinates from an ELK layout to be considered as an "Integer" (without digit after comma)?

[lredor]

Ask your questions
As discussed in forum today, I have a problem indirectly caused by the floating coordinates.

I have no simple reproduction use case. In ELK everything is OK. But after that, I "translate" ELK graph to a GMF graph and there are some glitches on edges (of one pixel). Because the source node have not the same absolute y location than the target node (because of coordinates with xxx.5)

Reproduction use case files
The following zip contains three ELKT files:

• 1-beforeELKLayout.elkt : The original diagram.
• 2_afterELKLayout.elkt : The diagram after the ELK layout.
• 3_manualSimulationOfResultOnGMFSide.elkt : The same diagram manually modified to observe the GMT result.
  elktSampleFiles.zip

The graph seems to be too big to be sent with "ELK live" links.

Screenshots
The ELK result is like this:

The edge "toA1", between "C111" and "A1", is a straight edge.
The absolute y coordinate of C111 is 116.5+44.5+33+33=227.
The absolute y coordinate of A1 is 138+89=227.

But after conversion to GMF, C111 absolute y coordinate is 228 and the edge "toA1" has a bendpoint of one pixel.
I manually simulate the observed result in GMF. It is like this:

[lredor]

I finally found a solution on my side. The approach is to take into account the rounding of the parents' coordinates to avoid drifting too much (and therefore rounding up or down).

To try to clarify my approach, if I take the example of four nodes contained within each other, with each a location of {10.5, 10.5} relative to its parent: A / B / C / D.
The absolute coordinates on the ELK side are:

• A: {10.5, 10.5}
• B: {21, 21}
• C: {31.5, 31.5}
• D: {42, 42}

The result on the GMF side was:

• A: {11, 11}
• B: {22, 22}
• C: {33, 33}
• D: {44, 44}

Now the result on the GMF side is:

• A: {11, 11}
• B: {21, 21}
  *C: {32, 32}
  *D: {42, 42}

I don't know if there is really something to do on ELK side (except an option to directly round the coordinates). But with this trick, I have non longer my problem.