Building DataFrames.jl on windows requires admin privileges

[felipenoris]

This happens due to CodecZlib.jl .

See JuliaIO/CodecZlib.jl#43 .

[bkamins]

@nalimilan - I have wanted to remove readtable and writetable form the package sources for some time now. This is yet another argument for it (and it should make DataFrames.jl have lighter dependencies in general + load a bit faster on startup).

[nalimilan]

We could drop readtable, but it sounds quite easy to drop the additional dependencies, so... See #1755.

[davidanthoff]

Shouldn't this just be fixed in BinaryProvider? There seems to be an issue already: JuliaPackaging/BinaryProvider.jl#149.

[bkamins]

Yes - this is a long known issue. I simply think that we do not need these dependencies in DataFrames.jl anyway (in general my view is that DataFrames.jl should be 100% Julia including all its dependencies).

[bkamins]

Just to expand on for possible future readers of this thread:

DataFrames.jl should be 100% Julia including all its dependencies

We want to be sure that when DataFrames.jl fails badly (e.g. core dumps) this must mean only one of the following two reasons:

• there is a bug in Base
• an unsafe operation in-built into the language (like @inbounds) was used

so that the user can exclude the possible cause that e.g. some third-party C code used by DataFrames.jl has a bug.
This is in practice very relevant as columns of DataFrame can have a myriad of sources. People often report that DataFrames.jl fails. As of now - when we exclude the two points above by doing code analysis of DataFrames.jl functions only we can safely say that the bug is not in DataFrames.jl.