[GSoC] Dataframes for D
Prateek Nayak
lelouch.cpp at gmail.com
Tue Jun 25 17:25:34 UTC 2019
>>> [snip]
-------------
Week 4 Update
-------------
This marks the completion of Stage I of Google Summer of Code
2019. It seems like it was only yesterday when I started working
on this project and it has already been a month.
--------------------------
So what happened last week
--------------------------
* apply - to apply a function on a row/column
* function to convert a column of data to level of Index
* drop - to drop a row/column
Going back to the original proposal, I had allocated some time
for optimisations in case there was time:
I was testing old parser with large files and it failed
miserably. So I redesigned the from_csv function and added it to
the library as fastCSV.
fastCSV gives 40x speed improvement over from_csv and fastCSV
will eventually replace from_csv
* fastCSV was added to the library.
-------------------
Plans for this week
-------------------
Plans for this entire stage isn't strictly on a week by week
timeline but the following things will be dealt sequentially
throughout this stage:
This stage is reserved for implementation of groupBy. So for the
beginning the internal structure and grouping will be decided.
Later things like display and combining into a DataFrame struct
will be dealt with.
These tasks were scheduled for Stage-III but will again fall
under sequential implementation. If the above tasks are done. The
following tasks will be dealt with:
* Aggregate [with complete set of popular operations]
* Join will be implemented to merge two DataFrame.
Aggregate was reserved for later stages to support implementation
for both normal DataFrame and groubBy at once.
----------
Roadblocks
----------
This week there hasn't been any roadblocks. I needed the help of
my mentors to solve a couple errors here and there but other than
that things were smooth.
As for the future roadblocks, I cannot see any apparent ones but
then again they show up when you are least expecting them :(
More information about the Digitalmars-d
mailing list