doctest with variable return value

[3KWA]

Hi all,

I am wondering what is the standard doctest (test) practice for
functions who's returned value change all the time e.g. forex rate:

import urllib

def get_rate(symbol ):
"""get_rate(sym bol) connects to yahoo finance to return the rate of
symbol.

>>>get_rate('AU DEUR')

"""

url=
"http://finance.yahoo.c om/d/quotes.csv?s=%s =X&f=sl1d1t1c1o hgv&e=.csv" %
\
symbol
f=urllib.urlope n(url)
return float(f.readlin e().split(',')[1])

As you can guess I am very new to unittest and doctest in general ...

Thanks for your help,

EuGeNe

[Rob Sinclar]

On Tuesday 25 July 2006 09:53, 3KWA wrote:

Hi all,

I am wondering what is the standard doctest (test) practice for
functions who's returned value change all the time e.g. forex rate:

import urllib

def get_rate(symbol ):
"""get_rate(sym bol) connects to yahoo finance to return the rate of
symbol.

>>>get_rate('AU DEUR')

"""

url=
"http://finance.yahoo.c om/d/quotes.csv?s=%s =X&f=sl1d1t1c1o hgv&e=.csv" %
\
symbol
f=urllib.urlope n(url)
return float(f.readlin e().split(',')[1])

As you can guess I am very new to unittest and doctest in general ...

Thanks for your help,

EuGeNe

Hi EuGeNe,
Pass it through a variable before returning a value.
Here's how I would do it:

import urllib2
def get_rate(symbol ):

URL='http://finance.yahoo.c om/d/quotes.csv?s=AU DEUR=X&f=sl1d1t 1c1ohgv&e=.csv'
request_headers = { 'User-Agent': 'Linuxinclar/0.1' }
request = urllib2.Request (URL, None, request_headers )
response = urllib2.urlopen (request)
STR = response.read()
return STR.split(',')[1].strip()

SYMB='AUDEUR'
print SYMB,'=',get_ra te(SYMB)

Python rocks.
That's be nice to indicate hour though (4th array element)...

Best Regards,
Rob Sinclar

[Peter Otten]

3KWA wrote:

I am wondering what is the standard doctest (test) practice for
functions who's returned value change all the time e.g. forex rate:

import urllib

def get_rate(symbol ):
"""get_rate(sym bol) connects to yahoo finance to return the rate of
symbol.

>>>get_rate('AU DEUR')

"""

url=
"http://finance.yahoo.c om/d/quotes.csv?s=%s =X&f=sl1d1t1c1o hgv&e=.csv" %
\
symbol
f=urllib.urlope n(url)
return float(f.readlin e().split(',')[1])

You cannot test for an unknown value, but you can do some sanity checks:

>>rate = get_rate('AUDEU R')
rate 0

True

>>isinstance(ra te, float)

True

This will at least make sure that get_rate() does not throw an exception.
You can also spoonfeed it with handcrafted data...

>>def mock_urlopen(ur l):

... from cStringIO import StringIO
... return StringIO("yadda ,0.1234")
...

>>urllib.urlope n = mock_urlopen
get_rate("AUD EUR")

0.1234

but this has the disadvantage that the test has to know about the actual
implementation of the function about to be tested.

Peter

[3KWA]

Peter Otten wrote:

You cannot test for an unknown value, but you can do some sanity checks:

>>rate = get_rate('AUDEU R')
>>rate 0

True

>>isinstance(ra te, float)

True

This will at least make sure that get_rate() does not throw an exception.

Thanks a lot ... sanity checks ... it makes a lot of sense to me!

At EuroPython I attended a talk where someone said that untested code
is nothing ... so I am trying to write something instead of nothing ...
on the other hand can code ever be over tested?

EuGeNe

[Duncan Booth]

3KWA wrote:

>
Peter Otten wrote:

>You cannot test for an unknown value, but you can do some sanity
checks:

> >>rate = get_rate('AUDEU R')
>>rate 0

True

> >>isinstance(ra te, float)

True

This will at least make sure that get_rate() does not throw an
exception.

Thanks a lot ... sanity checks ... it makes a lot of sense to me!

At EuroPython I attended a talk where someone said that untested code
is nothing ... so I am trying to write something instead of nothing
... on the other hand can code ever be over tested?

Yes, code can be over tested: tests require maintenance just like any other
code, so you should avoid having tests which just duplicate other tests and
don't add any value. e.g. If you know the rate code works for a few
currencies it probably also works for most others, so you don't need to
exhaustively test all possible currency pairs.

For your unknown value, I would choose a range which you might expect to
hold true for a reasonable period of time and check the value is inside
that range. If the currencies shift massively you might need to update the
test, but with luck that won't happen:

>>0.4 <= get_rate('AUDEU R') <= 0.8

That way if your code starts accidentally returning EURAUD you should catch
it in the test, but minor shifts shouldn't matter.