use DataFrames as input for performance_profile()

[lrsantos11]

Using data frames from DataFrames.jl as input for performance_profile() would be usefull.
Perhaps, the syntax could be to have a DataFrame as

df = DataFrames(ProbName=String[],Solver1=Int[],Solver2=Int[],Solver3=Int[])

then push to df the data from each solver and call

performance_profiles(df)

which would use the column name from df as name of the solver, if note given by the user.

[tmigot]

I believe this is very similar to performance_profile in SolverBenchmark.jl, where the input is stats::Dict{Symbol,DataFrame} whose keys are solver names and values data frames from the solver call.

[lrsantos11]

Sure!

[dpo]

@lrsantos11 What would be the advantage of pushing data into a DataFrame as opposed to an array? The function in SolverBenchmark extracts data from a DataFrame into an array...

[lrsantos11]

I understand that many have been keeping their results using DFs.

Perhaps moving or copying the performance_profile function in SolverBenchmark.jl as mentioned by @tmigot as it seems to be the solution I was looking for. The advantage is that one do not need to call another package to use this feature.

I thinkg there is no performance (no pun intended) improvement using DFs whatsoever, yet usability will be improved, in my opinion.

[dpo]

I don’t see any problems with that. @abelsiqueira, how about you?

@lrsantos11 do you want to submit a PR to each repo?

[abelsiqueira]

I'm happy with it too.