Conditional sum on edges in julia - julia

Question

I have a complex problem where I have to evaluate a sum on edges from a graph previously built.
Julia doesn't seem to handle conditional sum with edges.
Here is a simple problem similar to the one I am trying to solve:
module EssaiModule
using LightGraphs, MetaGraphs
g = DiGraph(6)
mg = MetaDiGraph(g, 1.0)
add_vertex!(mg)
add_edge!(mg,1,2)
add_edge!(mg,1,3)
add_edge!(mg,1,4)
add_edge!(mg,2,4)
add_edge!(mg,2,5)
add_edge!(mg,3,5)
add_edge!(mg,5,6)
add_edge!(mg,4,6)
set_props!(mg,3,Dict(:port=>1,:vessel=>2))
set_props!(mg,1,Dict(:port=>1,:vessel=>0))
set_props!(mg,2,Dict(:port=>1,:vessel=>0))
set_props!(mg,4,Dict(:port=>1,:vessel=>2))
set_props!(mg,5,Dict(:port=>0,:vessel=>2))
set_props!(mg,6,Dict(:port=>0,:vessel=>0))
SI = sum(1 for e in edges(mg);get_prop(g,dst(e),:vessel)==2 && get_prop(g,dst(e),:port)==1)
println(SI)
end
When I test it like that I get the error
#LoadError: MethodError: no method matching dst(::Irrational{:e})
I really need to figure out how to sum with conditions because in my real problem I am putting this sum in a constraint, like this with the sum on edges ed and x a variable:
#constraint(model, cM[e in edges(g)], x[e] + sum(x[ed] for ed in edges(g) ; fn1(ed) == 2 && fn2(ed) == 1) <= 1 + y)
and I have the error message saying
#LoadError: UndefVarError: ed not defined
So my questions are the following:
Why is the first code not working, since the type of e is Int?
Am I writing my constraint in a bad way?
I am sorry I can't upload the real problem, the content and the data are not to be put online, plus it is too big.

Answer 1

For the first part of your question, you need to change your function call
SI = sum(1 for e in edges(mg);get_prop(g,dst(e),:vessel)==2 && get_prop(g,dst(e),:port)==1)
# ^ you end the statement with a semicolon
# after the semicolon, keyword arguments of a function, in this case `sum`,
# begins. then, the compiler checks the function `dst` and cannot find a proper
# method that is implemented for e, that is, the irrational number e.
to read as below:
SI = length([e for e in edges(mg) if get_prop(mg, dst(e), :vessel) == 2 &&
get_prop(mg, dst(e), :port ) == 1 ])
println(SI) # prints 3
Note also the typo in your code. get_prop needs to check mg, instead of g. You should also have a look at the comprehensions section of the documentation.
For the second part of your question, you should provide the macro definition for us to help you better. Most likely, you are not interpolating the expression/variable inside the macro. See my answer to a different question and tell if that helps.
EDIT 1. Actually, you have the same problem in your second part, too. Inside sum, you end the positional arguments with the semicolon and start the keyword arguments.
EDIT 2. As per DNF's comment below, you should even use count for a more expressive (in source code) and efficient version:
SI = count(e->(get_prop(mg, dst(e), :vessel) == 2 &&
get_prop(mg, dst(e), :port ) == 1 ), edges(mg))
# or,
SI = count(get_prop(mg, dst(e), :vessel) == 2 &&
get_prop(mg, dst(e), :port ) == 1 for e in edges(mg))
EDIT 3. Performance comparison is below:
Pkg.add("BenchmarkTools")
using BenchmarkTools
#benchmark count(e->(get_prop(mg, dst(e), :vessel) == 2 &&
get_prop(mg, dst(e), :port ) == 1 ), edges(mg))
BenchmarkTools.Trial:
memory estimate: 7.73 KiB
allocs estimate: 53
--------------
minimum time: 2.201 μs (0.00% GC)
median time: 2.494 μs (0.00% GC)
mean time: 3.191 μs (15.69% GC)
maximum time: 198.515 μs (95.69% GC)
--------------
samples: 10000
evals/sample: 9
#benchmark count(get_prop(mg, dst(e), :vessel) == 2 &&
get_prop(mg, dst(e), :port ) == 1 for e in edges(mg))
BenchmarkTools.Trial:
memory estimate: 7.75 KiB
allocs estimate: 54
--------------
minimum time: 2.352 μs (0.00% GC)
median time: 2.661 μs (0.00% GC)
mean time: 3.712 μs (15.10% GC)
maximum time: 236.440 μs (95.06% GC)
--------------
samples: 10000
evals/sample: 9
#benchmark length([e for e in edges(mg) if get_prop(mg, dst(e), :vessel) == 2 &&
get_prop(mg, dst(e), :port ) == 1])
BenchmarkTools.Trial:
memory estimate: 8.13 KiB
allocs estimate: 60
--------------
minimum time: 2.642 μs (0.00% GC)
median time: 2.789 μs (0.00% GC)
mean time: 3.553 μs (14.54% GC)
maximum time: 231.424 μs (94.73% GC)
--------------
samples: 10000
evals/sample: 9
count version seems to be 10% more efficient than length, which can make a difference in your actual, bigger problem.
EDIT 4. The proper way to do the benchmarking is below, thanks to DNF, again:
#benchmark count(e->(get_prop($mg, dst(e), :vessel) == 2 &&
get_prop($mg, dst(e), :port ) == 1 ), edges($mg))
BenchmarkTools.Trial:
memory estimate: 7.72 KiB
allocs estimate: 52
--------------
minimum time: 2.320 μs (0.00% GC)
median time: 2.478 μs (0.00% GC)
mean time: 2.952 μs (10.33% GC)
maximum time: 117.306 μs (93.38% GC)
--------------
samples: 10000
evals/sample: 9
#benchmark count(get_prop($mg, dst(e), :vessel) == 2 &&
get_prop($mg, dst(e), :port ) == 1 for e in edges($mg))
BenchmarkTools.Trial:
memory estimate: 7.73 KiB
allocs estimate: 53
--------------
minimum time: 2.340 μs (0.00% GC)
median time: 2.524 μs (0.00% GC)
mean time: 3.030 μs (11.17% GC)
maximum time: 197.018 μs (94.35% GC)
--------------
samples: 10000
evals/sample: 9