a possible fix to historical quotes - #48
Conversation
|
so works in python 2.7 but not 3.x |
| content = str(resp.read().decode('utf-8').strip()) | ||
| daily_data = content.splitlines() | ||
| content = resp.read() | ||
| quotes = re.findall('{"date":\d+[^}]+}', content) |
There was a problem hiding this comment.
Here you would need to call .decode() on content before passing it to re.findall.
It seems that you would also have to check if it's Python 3 and call .decode() only if it's Python 3. There may be other, more elegant, way of doing this -- I'm not used to port Python code.
There was a problem hiding this comment.
yes adding .decode('utf-8').strip() seems to get past this
| for quote in quotes: | ||
| j = json.loads(quote) | ||
| for k in ('open', 'close', 'high', 'low', 'unadjclose'): | ||
| j[k] = "%.2f" % j[k] |
There was a problem hiding this comment.
ystockquote.get_historical_prices('GOOGL', '2006-01-01', '2017-05-29') would throw KeyError: 'open' at this line.
There was a problem hiding this comment.
starforever I do not get that error either in my own tests or running the test suite.
(note I have not tested in Python 3.x)
Does it happen in Python 3.x?
There was a problem hiding this comment.
I'm using Python 2.7. I'm not sure if you are using the exact same arguments. It only happens for some specific symbol and time ranges. I haven't got time to dive deep on this. So no idea how it happens.
There was a problem hiding this comment.
I'm guessing it doesn't throw the error in the tests since the specific arguments are not used in tests. I silently caught the exception and it worked fine then. But I think this is only a quick fix.
There was a problem hiding this comment.
ok I see the issue, when the quote/date range includes a split or dividend the pattern match pulls it in but there is no 'open', 'close' etc.
I think I may add to the tests a request that includes a split and will hard code some values to also check for the adj/un adj case you noted in another comment
There was a problem hiding this comment.
yahoo's current data does not include un-adjusted prices anymore
All prices are adjusted for splits
I had to stop using yahoo a while ago b/c of this
|
|
||
|
|
||
| __version__ = '0.2.6dev' # NOQA | ||
| __version__ = '0.2.7dev' # NOQA |
There was a problem hiding this comment.
this should still be 0.2.6dev.. next release will be 0.2.6 (assuming we can fix this) .... and afterwards, 0.2.7dev will open
| from urllib2 import Request, urlopen | ||
| from urllib import urlencode | ||
| from datetime import datetime, timedelta | ||
| import re |
| content = str(resp.read().decode('utf-8').strip()) | ||
| daily_data = content.splitlines() | ||
| content = resp.read() | ||
| quotes = re.findall('{"date":\d+[^}]+}', content) |
There was a problem hiding this comment.
yes adding .decode('utf-8').strip() seems to get past this
| d = timedelta(seconds=j["date"]) | ||
| # these keys are for backwards compatibility | ||
| for key in j.keys(): | ||
| j[key.capitalize()] = j[key] |
There was a problem hiding this comment.
im getting an error running the tests on this line. not sure what you are trying to do here
| # these keys are for backwards compatibility | ||
| for key in j.keys(): | ||
| j[key.capitalize()] = j[key] | ||
| j['Adj Close'] = j['unadjclose'] |
There was a problem hiding this comment.
unadjclose means the real price, while Adj Close should be adjusted price w.r.t. splits and dividends. Assigning one to the other would lead to confusion and potential incorrect results.
There was a problem hiding this comment.
thanks starforever that is a straight up bug.
The issue is w/ the api change "close" used to be the unadjusted close but now it is the adjusted close.
I will adjust shortly.
from cgoldberg: "this should still be 0.2.6dev.. next release will be 0.2.6"
cgoldberg: "alphabetize imports says PEP8"
use the lower case keys to get the "new" values Not sure if under the old api "High" / "Low" was adjusted or not If so more work will need to be done to be backwards compatible
5 participants
Since things broke when the API changed I wrote this.
It relies on the fact that the web page has all the data in json objects.
So I pattern match to grab the json fragments for the quotes (so a bit brittle).
I do this instead of the download link b/c the download link is generated via javascript and requires cookie handling... so while the data is better formatted getting to it requires reverse engineering, so brittle in a different way.