Search Results | chat.stackexchange.com

RobotHumans

Jul 24, 2018 20:02

case in point example is when i fixed pandas-datareader for edgar, and the PR got denied. the problem is still there. or fixed w3af deployment on macosx by using docker. again pr denied.

RobotHumans

May 30, 2017 08:24

probably have to import it explicitly now. version mismatch between keras and pandas.

RobotHumans

May 5, 2017 00:47

@SergiyKolodyazhnyy 30 or so of mine were for work on a private repo. 2 were on my own project. one was to fix a feature for pandas-datareader and put in a PR, but I only fixed it on py3k, so until someone steps in and forks my repo to touch it up for py2x, the PR won't get approved.

RobotHumans

Apr 29, 2017 21:02

I'm pegged at 100% right now, but it's because I restarted my pandas/talib precompute store b/c with the upgrade to pandas 0.19 rounding rules changed, and I want it homogenous.

RobotHumans

Apr 7, 2017 01:16

I also have a pending PR in to pandas-datareader from the spare time tinkering. passes CI on pandas 3 and all supported versions of pandas. Have to get it to pass on 2.7 for three different versions of pandas.

RobotHumans

Mar 31, 2017 22:09

nice. I was goofing earlier on a pandas project. back to work stuff between now and Tues though.

RobotHumans

Mar 31, 2017 20:56

and a 3 month old bug in part of pandas got bonked today. ;)

RobotHumans

Mar 22, 2017 19:10

getting close to time to slick my box. i was running pandas .13 and it's up to .19 now, and i'm in dependency hell

RobotHumans

Mar 10, 2017 22:33

@terdon yeah, i'm in here a lot less. I'm in a stock trading discord most of the time now. I wrote that python pandas talib layer thing, and I'm playing with quantopian. If you refine existing strategies to be better than the rest, they fund your strategy with 1-5 mil and give you 10% of gains. So, if nothing else, an interesting hobby.

hbdgaf

May 19, 2016 00:20

I'm thinking about revisiting pistols for pandas and using some of the machine learning stuff i'm studying now.

hbdgaf

May 10, 2016 03:52

I'm looking forward to snaps for things with tons of dependencies like pandas or scikit-learn, so you can run head from git without making your system less reliable.

hbdgaf

Mar 9, 2016 02:48

I think I found a few decent ones on R and pandas that were mirrored on github.

hbdgaf

Feb 14, 2016 01:42

R is interesting, but I'ld be a lot more interested if it were using pandas or numpy and some graphing lib.

hbdgaf

Dec 30, 2015 02:37

I was thinking that old pistols for pandas project I'm still tinkering with. All those precompute tables.

hbdgaf

Jul 19, 2015 18:43

using the ORM is slower if you count time to ispect pandas frames and programmer time to write the perfect algo. it's faster to eat the time on sqlkit commits then introspect the database. so they're both losses and wins in their own way.

hbdgaf

Jul 6, 2015 00:58

@Mateo definitely right. I wouldn't get loud about it... just sort of a "Six videos on fetching data from yahoo? how about from pandas import web. web.fetch(YAHOO, FLWS). Have some flowers"

hbdgaf

Jul 6, 2015 00:55

@RPiAwesomeness github.com/aking1012/pandastalib - using pandas and flask to precompute all the things in talib and make it browseable/searchcable offline at your leisure. also adding some additional precomputes and features.

hbdgaf

Jul 6, 2015 00:54

looking at to_dict for a pandas frame and moving closer to core on flask to get it done. as of flask 0.10 that looks like the right thing to do, but it means i have to use pip instead of ubuntu packages. pandas and flask are significantly behind on LTS

hbdgaf

Apr 24, 2015 00:46

I've been working on a layer for pandas and talib to do all the technical analysis stuff as precomputes. Find a target set. Do some level of stastical favorability for each TA brick. Then restack them over and over until you find the best set of conditions yet. Repeat ad nauseam ad infinitum. Different goal. A lot of the same tools. The precompute bit is to help the computer learn which conditions are most favorable.

hbdgaf

Apr 20, 2015 22:34

@Mateo - did I tell you I have the pandas frame/csv/precompute/sync/cleaning-csv/cleaning-frame-dupe-columns/slamming-in-t‌o-database-with-flask-sqlalchemy parts done now?

hbdgaf

Apr 18, 2015 17:23

Anyway... working on the CSV from pandas to SQL layer. Making substantial progress, but it's making me realize how sloppy the work was so far.

hbdgaf

Feb 7, 2015 00:06

think of it this way, if i get this working the way i want, i'll definitely be donating to pandas/talib. if i make money, they make money. it's just RIGHT.

hbdgaf

Jan 31, 2015 21:26

at least they were happy pandas instead of sad pandas...

hbdgaf

Jan 12, 2015 23:29

i'm working on fixing talib with an abstraction layer, so now you can basically throw a pandas df at talib and it just saves the precomputed junk in a csv file. now you just have to join frames to get at whatever you want. so, it's a small bit odd... but it's a solution to a problem.

hbdgaf

Jan 10, 2015 21:28

Taking the talib abstract api and making a sort of organized api that you just throw a pandas dataframe at, and it returns a df with just the results or optionally appends it to the existing frame. That way you just say "compute all of everything talib knows about for this historical data(blob)" and it just saves it for you.

hbdgaf

Jan 3, 2015 23:48

other news, pandas 0.15 added an options info api. that should prove useful if i ever get this project to work... it's one of those "each individual part is easy, but there are SOOOO many parts" things.

hbdgaf

Jan 3, 2015 16:12

So, I finished the "candlestick patterns precompute" part of that thing. Still slowly plugging away at the rest of the pre-compute tables, but one of the guys that's using talib professionally liked the layer I'm putting in between pandas and talib enough to star it on github. Nice to know it's not a horrible idea I guess.

hbdgaf

Dec 23, 2014 01:29

I just need to remember to be awesome and keep writing that pandas/talib layer.

hbdgaf

Dec 19, 2014 00:14

thinking about asking mrjbq7 and the pandas community if they'ld consider a pull to one place or the other to make that little bit of "wow that was easy" to the rest of everyone.

hbdgaf

Dec 16, 2014 21:36

@NathanOsman nice. i recently got pandas, talib, and flask playing nicely on 3.4, but i'm having trouble with the flask-openid package not doing the appropriate redirects or prompts for logins... that's my next hurdle.

hbdgaf

Nov 25, 2014 20:16

nice. i'm playing with using anaconda to get all my deps right for some python and getting pandas up to not ancient versions so i can use postgres instead of csv

hbdgaf

Nov 23, 2014 00:47

i know i want edge pandas and matplotlib without having to pay the just use arch or compile from source and fight dependency hell thing.

hbdgaf

Nov 13, 2014 21:43

oh, it's totally fine. i'm trying to make it stupid simple. here goes:

You've got your historical data set. You want to reduce it to biggest movers up or down by percentage and test it historically. So, first you do all the "this could refine the set" logic bricks to whittle down pandas frames.

In order to visualize how correct or incorrect each brick indicates, you set a pixel/block for each item you're tracking to a color, say red for down and green for up. Each found and is right historically adjust the red or green RGB value. A misfire adjusts the B value.

hbdgaf

Nov 9, 2014 01:32

So... finally getting not as terrible and pandas data manipulation... and a little better at matplotlib usage. iPython is a lifesaver though.

hbdgaf

Oct 26, 2014 18:38

I don't think it's a coincidence. IT people tend to have a thing for unicorns, pandas, and ponies.

hbdgaf

Oct 26, 2014 18:31

pandas the scipy layer not a restaurant

hbdgaf

Oct 25, 2014 21:24

more of the same... playing with pandas at the moment.

hbdgaf

May 25, 2014 21:38

@Lucio i see... i'm playing with pandas at the moment though

hbdgaf

May 25, 2014 06:54

@NathanOsman cool. i guess i have to test it myself. i want to access raw data from pandas with read and to_sql. (so it's not really raw sql) I was just wondering if auto-keys would just work when I write-back with pandas/scipy

hbdgaf

Apr 21, 2014 20:50

so, the less awful i get at pandas, the more i can't understand how i didn't get it doing what i wanted sooner.

hbdgaf

Apr 15, 2014 23:06

well this should make my life easier... github.com/chrisdev/django-pandas

hbdgaf

Apr 10, 2014 22:03

later on people. more playing with pandas.

3

hbdgaf

Apr 8, 2014 21:09

I'm so convinced pandas+scipy is awesome, but I'm not using it right yet...

hbdgaf

Apr 7, 2014 23:00

back to pandas and complex math.

hbdgaf

Feb 22, 2014 22:46

@Seth R is interesting, but i'm more a fan of pandas or scipy