Closed dmbates closed 8 years ago
I was wondering about this too. +1
I'm for this, too.
The only downside I see is that in RDatasets.jl, DataFrames with legitimate categorical columns (not just because strings were often pooled in R) are stored as .rda files.
Maybe move the file into RDatasets.jl in the short term and move those files to feather or .jld longer term? c.c. @johnmyleswhite
At the very least, we could move read_rda
to RDatasets waiting for a larger refactoring.
+1
See also https://github.com/JuliaStats/DataFrames.jl/issues/958, which moving to RCall.jl would likely fix.
@alyst What's the state of your RData package? Is it such that we could remove the RDA stuff in here entirely and direct users to your package?
For the others, see https://github.com/alyst/RData.jl
@ararslan I'm using it regularly and it works for me. It's an extended version of DataFrames.read_rda()
that supports all R data types (so it does not crash on an arbitrary RData file, although not every R type is converted to Julia counterpart). So, yes, I think you can recommend it to the users. I suspect the first users might run into issues, because RData
received limited testing so far, but I will try to fix the bugs as they are discovered.
Great. So let's deprecate RDA.jl ASAP in favor of RData.jl. A good stress test for it is porting RDatasets.jl and checking that all data sets can be loaded correctly.
First RData needs to be registered in METADATA. Should also set up Travis and coverage.
@alyst If you want you could transfer ownership of RData.jl to the JuliaStats organization for more rapid development and maintenance, but that's totally up to you.
@ararslan If you think that making RData
a part of JuliaStats would be beneficial for the community, I am all for it. Is there anything required from my side?
It would make sense to me to have it under the JuliaStats umbrella but I'd be interested to hear what others think. On your end you'd have to go to the repo settings and transfer ownership.
Bump. @alyst, are you still considering transferring ownership of RData.jl to JuliaStats?
@ararslan Yes I am, whenever JuliaStats "stakeholders" decide it's worth it.
Sounds like myself (though I'm hardly what you could call a "stakeholder") and @nalimilan (based on his thumbs up) think it's worth it. What do the other folks think? @johnmyleswhite?
+1
@alyst By show of thumbs (and Jacob), it looks like people are thinking it's worth it. 😄
@ararslan Great! I've tried to transfer RData.jl
to JuliaStats, but got an error saying that I'm not an admin of JuliaStats to make the transfer. I've never transferred repos before, so I don't know whether JuliaStats admins have received notifications about pending transfer and can complete it, or there is no other way than to make me an admin temporarily.
@alyst: You have the necessary permissions now.
I've just transferred the repo. Have fun!
Thanks @alyst!
Can an owner set the RData permissions so that JuliaStats members have our usual permissions? When doing transfers on GitHub the permissions are copied over too, so regular members don't currently have commit access.
@ararslan I've just added the same permissions as DataFrames.jl
For some reason I'm not able to enable Travis CI on it. I've been able to do it with other repos... not sure what the issue is.
@ararslan TravisCI needs to be enabled in the settings (currently it's not), but since it requires some configuration, it's better that the more experienced admins set it up.
RData package is now registered
This code to read a subset of the
.rda
(R data archive) format was used to import saved Rdata.frame
objects. TheRCall
package is much better at this. I suggest deprecatingread_rda
while doing the conversion for julia v0.5 I believe this will also remove the dependency on theGzip
package.