Speed Benchmarks

[Erotemic]

I was looking at benedict's docs and implementation and it caught my eye that it is a drop-in replacement for dict. But I was wondering if the speed impact of doing this has been measured.

For context, I was comparing my own dictionary subclass: ubelt.udict, which I also use as a drop-in for dict QOL enhancement. I was impressed with what benedict had, and I was wondering what the slowdown cost of using it is.

I did an initial experiment where I took one case from a previous benchmark I wrote, and dropped-in a benedict.BeneDict, ub.udict, and dict, and there seems to be a significant slowdown for benedict.

Did I hit a bad test-case? Is the expectation that you use benedict outside of code sensitive to the speed of dictionaries?

[fabiocaccamo]

@Erotemic thank you for your benchmark! As you imagined, my mainly benedict usage is more focused on simplifying common operations than having code sensitive to the dictionary speed, in that case I would probably use a simple dict.

Said that, I don't want to say that I don't care about code speed, but it has not been the main focus of this library since the beginning.

If I should bet 2 cents on the possible cause, I think that the main one is the keypath_separator, probably disabling it will speed-up the code.

If you make any progress on this topic let me know, I would be glad to improve the performances of this library.