Performance issue while adding constraints

[Ervin66]

Dear all,

when googling for ways to speed up the constraints generation in a lot of old SO threads it is mentioned that using "+=" is awfully slow but I have yet to find what should be used instead. There was also mention of a patch for the "+=" operator has this been implemented?

If this has not been patched could this someone provide an example of what should be done instead? I think this questions has been asked by many but there hasn't been any satisfactory answer? example

Thank you all in advance

[stumitchell]

lpSum() and a list comprehension

so
c = lpSum([x for x in X])

[Ervin66]

but isn't that just summing over the element within lpSum?

if we take this example from the documentation how would this work?

for b in Bars:
    prob += lpSum([vars[w][b] for w in Warehouses])>=demand[b]

[pchtsp]

Do you have a specific minimum reproducible example in python we could check? Stuart's answer is correct. It just may be that the inefficiencies are coming from somewhere else in the code (or in the structure of the list comprehension).

In some very rare occasions, there are some very niche alternatives that can work better. That's why I'm asking for an example. And why there's no "single definitive solution" (that is, besides lpSum + list comprehension)

[Ervin66]

I am not sure whether I understand the list comprehension in this case. I already do a lsit comprehension in order to sum over the elements but how could I use it to generate all the constraints and then only add them later?

for b in Bars:
    prob += lpSum([vars[w][b] for w in Warehouses])>=demand[b]

How would you iterate over the Bars list in this case in order to add all constraints at once?

[stumitchell]

@Ervin66 you are currently doing things right.
The suggestion you refer to is there because often people come to pulp and generate code like the following

for b in Bars:
    c = LpAffineExpression()
    for w in Warehouses:
        c += vars[w][b]
    prob +=c >=demand[b]

in this case the += in this line c += vars[w][b] makes the code very slow

[stumitchell]

Pulp uses the += operator to add constraints to LpProblems this is idiomatic usage.

However often people come from other libraries and languages and use += to generate single constraints like the following

for b in Bars:
    c = LpAffineExpression()
    for w in Warehouses:
        c += vars[w][b]
    prob +=c >=demand[b]

in this case the += in this line c += vars[w][b] makes the code very slow

[Ervin66]

thanks for the reply and help! :)

But for my objective function I create a lpSum element for each individual summation and once they are all created I add to the model at once which does speed the creation of the objective function example:

c1 = lpSum[(x[i][j]*c[i][j],
                 for i in I,
                 for j in J])
c2 = lpSum([Y[i]*fc[i],
                  for i in I])

prob += c1 + c2

So there is no equivalent for constraints to be all added simultaneously?

[stumitchell]

LpProblem.extend() will do what you want https://github.com/coin-or/pulp/blob/b99ff24891123fae928c6571d2eccd3e073eb385/pulp/pulp.py#L1681 Stuart Mitchell PhD Engineering Science Extraordinary Freelance Programmer and Optimisation Guru www.stuartmitchell.com

…

On Tue, May 4, 2021 at 4:06 AM Ervin Python ***@***.***> wrote: thanks for the reply and help! :) But for my objective function I create a lpSum element for each individual summation and once they are all created I add to the model which does speed the create of the objective function example: c1 = lpSum[(x[i][j]*c[i][j], for i in I, for j in J])c2 = lpSum([Y[i]*fc[i], for i in I]) prob += c1 + c2 So there is no equivalent for constraints to be all added simultaneously? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#436 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAFUIWLQ5WZQABU5AOYFXHTTL3CW7ANCNFSM432DHM7A> .

[Ervin66]

I will look into it, thank you stu