I made a small change to how the Diagram uses intermediate legs with mp mode. Such legs must be used in XBetza as ''seemless glue' for combining multiple legs into a single leap, for making that leap commensurate with a later leap in the path. This because XBetza must express all legs of a multi-leg move with the same atom.
E.g. the Grant Acedrex Unicorno, moving as N-then-B, would need an additional mpW step to start with ([mpW-B]), and express the whole path with steps of the K atom. The mp makes this first step completely insensitive to the occupancy of the W squares, as an N leap should be.
The XBetza compiler now merges such seemlessly glued legs into a single one. This is a speed optimization, but it has effect on the way a type-specific hopping ban in the captureMatrix is applied. Formerly this was also applied to the mp legs, such that a Unicorno hopper that was forbidden to hop over type X would also be blocked by an X on the W squares when it was described as [N?pB]. I.e, the initial step would behave as a Mao. I considered this undesirable, which was the main reason to change the behavior. A ski version of the Korean Cannon should not be blockable on an adjacent square.
Note that a complete Korean Ski-Cannon would be [D-pR][pD-R] to also hop over a mount on the D squares. If it was desirable to make it blockable by another Cannon on the W squares one can always split the m and the p part of the dummy intermediate leg of the D leap as mW-W plus pW-W. By making the hop onto the W square explicit, and not combined with m, it would again be under control of the captureMatrix.
I made a small change to how the Diagram uses intermediate legs with mp mode. Such legs must be used in XBetza as ''seemless glue' for combining multiple legs into a single leap, for making that leap commensurate with a later leap in the path. This because XBetza must express all legs of a multi-leg move with the same atom.
E.g. the Grant Acedrex Unicorno, moving as N-then-B, would need an additional mpW step to start with ([mpW-B]), and express the whole path with steps of the K atom. The mp makes this first step completely insensitive to the occupancy of the W squares, as an N leap should be.
The XBetza compiler now merges such seemlessly glued legs into a single one. This is a speed optimization, but it has effect on the way a type-specific hopping ban in the captureMatrix is applied. Formerly this was also applied to the mp legs, such that a Unicorno hopper that was forbidden to hop over type X would also be blocked by an X on the W squares when it was described as [N?pB]. I.e, the initial step would behave as a Mao. I considered this undesirable, which was the main reason to change the behavior. A ski version of the Korean Cannon should not be blockable on an adjacent square.
Note that a complete Korean Ski-Cannon would be [D-pR][pD-R] to also hop over a mount on the D squares. If it was desirable to make it blockable by another Cannon on the W squares one can always split the m and the p part of the dummy intermediate leg of the D leap as mW-W plus pW-W. By making the hop onto the W square explicit, and not combined with m, it would again be under control of the captureMatrix.