Adding ports to connected components can change connections

[fpw]

Adding ports to connected components can change connections

[flandreas]

@fpw The reason why this happens is that Antares doesn't store the port names when connecting wires to subcircuit pins. This is because it would break when renaming pins. Instead, it uses the internal index of the port, such as 1, 2, 3 and so forth.

Now, until a couple of releases ago, this principle worked fine even when adding additional ports to a subcircuit. Because the user had to manually add these ports to the symbol, the additional ports received higher port numbers, and the existing one were untouched.

This changed when we introduced the "Generate Symbol" feature. A symbol is automatically laid-out again whenever a port is added or removed. The algorithm first removes all ports, then adds all inputs to the left side and all outputs to the right side of the symbol, leading to new port IDs. This is why in your example, the LED, which was previously connected with former port 2 (the O1 output), is now connected to I2, because that is now the port 2 (because it is the second port after the relayout.

I have to think about how to change the layout algorithms to make the port assignments more robust with these use cases.

[fpw]

Thanks for the explanation of the internals! Not sure if it helps, but my first idea would be this:

• modify the "Generate Symbol" algorithm to optionally get a mapping from name to id as input
• when a port is added to a "Generate Symbol" circuit, retrieve the name-to-id mapping of the current symbol
• delete all ports (like now)
• call the "Generate Symbol" algorithm with the mapping retrieved in step 2
• now the algorithm can assign old IDs to old names and use new IDs for ports that aren't in the old mapping (i.e. the new ports that were just added)
• create only those ports that appear in the new assignment, using the former port IDs from the table

[flandreas]

@fpw Yes that helps, I'll do it along these lines.