dalley has quit [Remote host closed the connection]
dalley has joined #hpy
dalley has quit [Ping timeout: 268 seconds]
dalley has joined #hpy
dalley has quit [Remote host closed the connection]
dalley has joined #hpy
mkoubaa has quit [Ping timeout: 276 seconds]
mkoubaa2 has quit [Ping timeout: 276 seconds]
hexology has quit [Quit: hex on you ...]
hexology has joined #hpy
<steve_s>
Re NumPy and heap types migration. One remaining thing to deal with is how to access the heap types instances. With static types you have a global C variable. Heap types should be, in theory, passed around in module state. We can do that and then add module state support to HPy (https://github.com/hpyproject/hpy/pull/328), but at the same time we may have better alternatives: HPyGlobal and/or some module context thingy (outlined in
<steve_s>
that PR too). We can also just convert to heap types, but still store them in global variables and do the transformation to some per-interpreter HPy specific solution later. That would also make the change smaller.
<mattip>
+1 for eventually supporting module state, it is "the way of the future". But if it is easier for now to do a work-around that is fine too
<antocuni>
yes, +1 to migrate to heap types first, and then find a solution for the module state later