Feature Request: Compiler API

[sirwhinesalot]

Right now when working with PyCSP3, you create one model in a script, which is then compiled to XCSP3 when executed.

I would much rather have the option to script the actual compilation process, so I can create multiple different models through the PyCSP3 API, compile them through the API, and then solve them with whatever XCSP3 capable solver I choose.

Consider for example a single Jupyter notebook that models different constraint problems and wants to solve them separately.

The alternative is to generate the XML directly myself but that's not nearly as convenient as using the API, or to generate a python string in memory and then "exec" it to do the compilation.

[xcsp3team]

Currently, you can compile and solve instances in sequence (by calling clear() between each problem).
I am not sure that it is what you are looking.
Please, tell us.

An example below xith cells from a notebook:

from pycsp3 import *

x = VarArray(size=3, dom=range(3))

satisfy(
    AllDifferent(x)
)

allDifferent(list:[x[0], x[1], x[2]])

instance1 = compile()

print(instance1)

('ipykernel_launcher.xml', False)

ace = solver(ACE)
result = ace.solve(instance1)
print(result, values(x))

SAT [0, 1, 2]

result = ace.solve(instance1)
print(result, values(x))

SAT [0, 1, 2]

clear()
result = ace.solve(instance1)

 The instance has no variable, so the solver is not run.
Did you forget to indicate the variant of the model?

y = VarArray(size=3, dom=range(3))

satisfy(
    AllEqual(y)
)

allEqual(list:[y[0], y[1], y[2]])

instance2 = compile()
print(instance1 == instance2)
result = ace.solve(instance2)
print(result, values(y))

True
SAT [0, 0, 0]

[sirwhinesalot]

I didn't realize the compile function existed! I didn't see it in the docs...

But what I'd really like is for the "state" of PyCSP3 to not be global, meaning I can add new constraints to different instances, in the same execution, without "clearing".

csp1 = CSP()
x = csp1.VarArray(size=3, dom=range(3))

csp2 = CSP()
y = csp2.VarArray(size=5, dom=range(5))

csp1.satisfy(AllDifferent(x))
csp2.satisfy(...)

instance1 = csp1.compile()
instance2 = csp2.compile()

ace = solver(ACE)
ace.solve(instance1)
ace.solve(instance2)

# add more stuff to csp1
z = csp1.VarArray(...)
csp1.satisfy(...)
# etc.

[xcsp3team]

I will think about it soon and let you informed