Where now returns multiple matches - issue #2

AlexMili / torch-dataframe

Utility class to manipulate dataset from CSV file

MIT License

67 stars 8 forks source link

Where now returns multiple matches - issue #2 #6

Closed gforge closed 8 years ago

gforge commented 8 years ago

Ok, I think it works with the where issue. The changes became much bigger than anticipated as I stumbled upon a few more bugs, below is a list containing the main changes:

Now handles single element load/inserts
There are additional sanity checks for input size
The self.n_rows is now used for shape()
The shape() can now adequately handle table with nil elements
The insert now adds columns that previously weren't there
The _extract_row moved from private name scope to open and changed name to get_row

AlexMili commented 8 years ago

Wow amazing ! I did some minor comments but your work is really remarkable !

gforge commented 8 years ago

Great, thanks for the code-check, I've updated according to your suggestions but I'm not sure how you want to handle the filling with nan. Currently I assume a missing value if the index doesn't exist, should we instead force nan values?.

I also think I have a working solution for the value_counts and column_order - do you want me to add those branches to the current PR?

Btw - excellent work with adding the Travis support

AlexMili commented 8 years ago

Thank you for the adjustments !

Let's do it step by step. I am going to merge this one and I let you create different pull request for each issue. It will be cleaner and easier to follow each code review ;)

Concerning the NaN values let it like that, we'll see later if it's not convenient ^^