-
-
Notifications
You must be signed in to change notification settings - Fork 3
Conversation
At a glance, this looks good @jfb-h! Thanks for giving this a shot. Let me know how you found the process of figuring this out. Before merging this, I do have a few comments:
You can test writing to a dataset with Hope that helps 🙂 |
Thanks!
I tried to test the writer with julia> write(dataset("test"), df)
ERROR: TransformerError: There are no writers for "test" that can work with DataFrame
Stacktrace:
[1] write(dataset::DataSet, info::DataFrame)
@ DataToolkitBase ~/.julia/packages/DataToolkitBase/JUjFu/src/interaction/externals.jl:387
[2] top-level scope
@ REPL[10]:1 I tried to add a writer spec by including [[test.writer]]
driver = "arrow" in the Data.toml, but that didn't do the trick. The docs/tutorials don't seem to contain any info regarding writer specification, so maybe that's another thing that could be useful to add to the docs. I'd be happy to contribute a short writeup of my view of the involved steps once this is done, which maybe could inform a short tutorial for the docs. |
Thanks for the changes :)
❤️ |
Sorry for the delay, getting back to this now :) I'm looking into the writing, thing I've found a small error in Base. I should have a fix in the next day or few. |
Well, turns out it was easier than I thought: the blocking bug is fixed in tecosaur/DataToolkitCore.jl@c05ed5c and I've merged this PR with some modifications in 5e544cf 🙂 |
Perhaps it's a bit late after the fact, but if you have any thoughts on what documentation I could add to make it easier for other people to contribute/write their own data loaders, that would be great 😃 |
I'd be happy to! I'll try to write something up this or next week 🙂 |
Hi @tecosaur, I gave the arrow support a shot and here's what I got.
I tested the loader on a simple test dataset and it seems to work for
read(dataset("test"), Any)
andread(dataset("test"), DataFrame)
. I also implemented the writer but did not test it yet because I didn't know how, maybe you could take a look at that?Closes tecosaur/DataToolkit.jl#30