AutoViML
commented
5 years ago Hi Karkkainen: Thanks for your excellent comments. I will go over your suggestions carefully and get back to you. Would you mind trying It after I make some changes? Thanks again, Ram
On Fri, Aug 9, 2019 at 2:09 AM L. Kärkkäinen notifications@github.com wrote:
You need from autoviz import ... in the sample code. Preferably you should give a sample that can be just copy&pasted and run, and provide pictures of how it looks, so that one could evaluate whether to install this instead of the many other plotting libraries.
The dependencies are extremely heavy. Is it absolutely necessary to install Jupyter? Something inside also depends on sklearn, which was not included in pip deps.
As for CSV reading; if you are not able to autodetect/guess separators and date formats, do not bother "including" it in your library. It is just two lines of code to first load the data with pandas and then use another library for plotting, and in most cases one needs to do something in between anyway (data preprocessing).
An ideal plotting library would have API alike this:
from fictionalplot import Figure # if possible, keep it to just one simple import
fig = Figure() # Internally holds graphics context, Qt window, websocket to browser or whatever fig.plot(df) # display the graph and return instantly, try to auto-guess suitable format based on df
If using a Qt window, spawn a new process that does not terminate when the Python program ends, and that is automatically shared by all figures of all running programs (don't block execution of the program like Matplotlib does). If using Notebook/browser, you don't need separate process because browser already does that.
For true interactive plots (e.g. receive user input on scaling changes to recalculate new data in Python), use async/await to avoid blocking Python from executing while waiting for user input (but stay away from import asyncio which is utter crap -- instead use trio if you must).
Good luck with your plotting library. We could certainly use some good options (I am not entirely happy with either Matplotlib nor Plotly, and everything else is just bad).
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/AutoViML/AutoViz/issues/4?email_source=notifications&email_token=AMKBH6EOULYGKEYPF6QVEA3QDUCXPA5CNFSM4IKQ2ZC2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HEKDUXQ, or mute the thread https://github.com/notifications/unsubscribe-auth/AMKBH6A7DPQ55ZKUBLIFRBTQDUCXPANCNFSM4IKQ2ZCQ .
Tronic
You need from autoviz import ... in the sample code. Preferably you should give a sample that can be just copy&pasted and run, and provide pictures of how it looks, so that one could evaluate whether to install this instead of the many other plotting libraries.
The dependencies are extremely heavy. Is it absolutely necessary to install Jupyter? Something inside also depends on sklearn, which was not included in pip deps.
As for CSV reading; if you are not able to autodetect/guess separators and date formats, do not bother "including" it in your library. It is just two lines of code to first load the data with pandas and then use another library for plotting, and in most cases one needs to do something in between anyway (data preprocessing).
An ideal plotting library would have API alike this:
If using a Qt window, spawn a new process that does not terminate when the Python program ends, and that is automatically shared by all figures of all running programs (don't block execution of the program like Matplotlib does). If using Notebook/browser, you don't need separate process because browser already does that.
For true interactive plots (e.g. receive user input on scaling changes to recalculate new data in Python), use async/await to avoid blocking Python from executing while waiting for user input (but stay away from import asyncio which is utter crap -- instead use trio if you must).
Good luck with your plotting library. We could certainly use some good options (I am not entirely happy with either Matplotlib nor Plotly, and everything else is just bad).