add debugging support

[metaperl]

No description provided.

[s4w3d0ff]

After reading through the docs, mock would be appropriate for this repo, especially for unittests. I have no experiance with it, but it seems like exaclty what we want and it is part of the python3 standard library. I'd like to dig deeper and perhaps write a basic unittest using mock before merging these commits and having the repo depend on it.

[kdb424]

Try pytest for unit testing. It will simplify things and is considered one of the best in the industry.

[metaperl]

My comments here are a bit off-topic. My fork of python-poloniex has gone way away from the mainline and I dont expect all my changes to be accepted. One thing I would recommend though is a change in the way the none is produced. For instance, right now the nonce is assigned a value during object initialization: https://github.com/s4w3d0ff/python-poloniex/blob/master/poloniex/__init__.py#L112 then that value is bound into the arguments to be passed to the API: https://github.com/s4w3d0ff/python-poloniex/blob/master/poloniex/__init__.py#L182 and then the nonce is incremented defensively: https://github.com/s4w3d0ff/python-poloniex/blob/master/poloniex/__init__.py#L205 I would recommend that instead of nonce being an attribute of the object, that it be a property of the object as it is in my fork: https://github.com/metaperl/python-poloniex/blob/master/poloniex/__init__.py#L178 the advantage of this is that anytime it is requested, it automatically computes a fresh new value. You could get rid of the finally clause and also reduce the amount of initialization code by doing it this way. P.S. - Read about why I do not participate in the reference process <http://j.mp/refpolicy>. Terrence Brannon 818-359-0893 (cell)

…

On Wed, Mar 15, 2017 at 1:11 PM, Kyle Brown ***@***.***> wrote: Try pytest for unit testing. It will simplify things and is considered one of the best in the industry. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#56 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABTLeP3NPLK57JwU-TK9t6_AQKAvmH4ks5rmBvBgaJpZM4MTO8h> .

[s4w3d0ff]

@kdb424 pytest was going to be my first choice, but mock seems to be built for things like APIs. Again I know nothing about mock and how it actually works, but the docs claim:

It (mock) allows you to replace parts of your system under test with mock objects and make assertions about how they have been used.

And the main issue I am having with unittests is figuring out a clean way to bypass making real calls to poloniex.

Perhaps only make 2 test api calls to poloniex (1 private and 1 public) then the rest of the unittests would just be checking the url that would be sent to polo (without actually sending). This could probably be done without mock but if there is an easier way to do it with mock...

@metaperl The reason I have the nonce as a saved value is so that the nonce builds on itself (instead of calling the time() and int() functions every time), thus making execution of a new nonce take 1/3 the time it would of just getting a new number:

from time import time
from timeit import timeit

i = int(time()*1000)

def nonce():
    global i
    i += 1
    return i

def nonce2():
    return int(time()*1000)

timeit(nonce)
# 0.13286900520324707
timeit(nonce2)
# 0.3373839855194092

I am sure there is a better way to achieve this...

[metaperl]

On Wed, Mar 15, 2017 at 5:01 PM, s4w3d0ff ***@***.***> wrote: @kdb424 <https://github.com/kdb424> pytest was going to be my first choice, but mock seems to be built for things like APIs. Again I know nothing about mock and how it actually works, but the docs <https://docs.python.org/3.5/library/unittest.mock.html> claim: It (mock) allows you to replace parts of your system under test with mock objects and make assertions about how they have been used. And the main issue I am having with unittests is figuring out a clean way to bypass making real calls to poloniex.

Draw a distinction between unit tests and integration tests? The python BitTrex API has both types of tests: https://github.com/ericsomdahl/python-bittrex/blob/master/bittrex/test/bittrex_tests.py Perhaps only make 2 test api calls to poloniex (1 private and one public)

then the rest of the unittests would just be checking the url that would be sent to polo (without actually sending). This could probably be done without mock but if there is an easier way to do it with mock. @metaperl <https://github.com/metaperl> The reason I have the nonce as a saved value is so that the nonce builds on itself (instead of calling the time() and int() functions every time), thus making execution of a new nonce take 1/3 the time it would if getting a new number:

I see. However, I started getting rejections from Poloniex because of duplicate nonce values. I didnt save the exact error messages. But I re-implemented that part as a property and havent had the problem since.

…

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#56 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABTLfqNEY_9aJJHD_rOlTKHIY1icGs1ks5rmFGzgaJpZM4MTO8h> .

[metaperl]

P.S. - Read about why I do not participate in the reference process <http://j.mp/refpolicy>. Terrence Brannon 818-359-0893 (cell)

On Wed, Mar 15, 2017 at 5:01 PM, s4w3d0ff ***@***.***> wrote: @kdb424 <https://github.com/kdb424> pytest was going to be my first choice, but mock seems to be built for things like APIs.

`mock` is often used with a unit-testing framework. So you could use it with pytest, nose, etc However, using something like mock to override methods is debatable practice. Purists, and people from more strict OO backgrounds would say that properly designed OO software would be using Dependency Injection - https://github.com/google/pinject

[s4w3d0ff]

I see now how mock works, your right, it is probably better to just distinguish between the two types of tests.

I randomly came across this today: http://code.activestate.com/recipes/580745-retry-decorator-in-python/
I haven't tested it, but it seems similar to what your latest commit does. Could probably use it if you don't want to have another 3rd party lib to depend on. I am considering adding it to v0.3.x as an optional feature if it is indeed legit.

[metaperl]

Yes, I am getting exceptions from requests on a regular basis. So some sort of retry functionality is important. Here is what I put around the __call__ method today: metaperl@1076678#diff-eeaeffee5a0554a0261f80f7b06417c3R186 I personally like 3rd party modules because they get a lot of usage and a lot of patches from other people. Do you like manually pasting in code instead of 3rd party modules? P.S. - Read about why I do not participate in the reference process <http://j.mp/refpolicy>. Terrence Brannon 818-359-0893 (cell)

…

On Tue, Mar 21, 2017 at 3:14 PM, s4w3d0ff ***@***.***> wrote: I see now how mock works, your right, it is probably better to just distinguish between the two types of tests. I randomly came across this today: http://code.activestate.com/ recipes/580745-retry-decorator-in-python/ I haven't tested it, but it seems similar to what your latest commit does. Could probably use it if you don't want to have another 3rd party lib to depend on. I am considering adding it to v0.3.x as an optional feature if it is indeed legit. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#56 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABTLXd-1CwvpqmzuaA-LCEtd4I4G1fRks5roCF5gaJpZM4MTO8h> .

[s4w3d0ff]

I personally would not like to rely on 3rd party modules (I feel better about it if the module is adopted into python3 core libs), and since I have been advertising this repo as 'python2 and 3 compatible', I try to stay away from any unnecessary 3rd party modules that may cause more problems than they fix.

If all I need is a simple, single function from a 3rd party module, I would prefer not depending on it directly. If there is a need for more functionality from the module later on, then I consider depending on the entire module. Maybe it is just me... but it has saved me headaches.

This repo uses requests because it was removing LOTS of bugs (and does most of the work for you). With smaller 3rd party modules I have found them to change often (much like this repo).

I don't "like" to copy/paste code but if all I need is 20 lines of basic code, why download 20k+ lines of useless code onto your system to import 20 lines into another 3rd party module?