Re: [go-nuts] Re: How do I test an func which depends on (pseudo)-random numbers, e.g. rand.Intn()?

[Simon Ritchie]

As some posters have already said, it depends what you are trying to test.  

If you want to test what your solution does when it receives a 
randomly-generated number, then you don't need to use random numbers.  You can 
hide the generation of the random number in a type which is defined by an 
interface.  Then you can supply a dummy version of the type which supplies 
numbers that you choose specifically for your test.  So for example, if you 
want to test what your solution does when it receives a randomly-generated 
number less than 10, you set up a test that supplies it with 8.

On the other hand, if you want to test the random number generator, for example 
to test its randomness, then this approach is no good.

Simon

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: [go-nuts] Re: How do I test an func which depends on (pseudo)-random numbers, e.g. rand.Intn()?

[Michael Jones]

You could also capture a (p)random stream as a test data vector and use it
in lieu of the PRNG during tests. This is the same as seeding iff and only
iff the PRNG is never changed, Should that change, then so will the tests
unless you capture a test vector now. (For example, if you have a timing
benchmark and the amount of work varies, then having a a cache of static
random values makes the timing always be comparable.)

This is pedantic and probably not important to most people. But if it
matters to you, it is the answer.

On Wed, Mar 15, 2017 at 11:25 PM,  wrote:

> You would test it the same way you test anything that depends on IO, you
> don't make your function depend on a random number or an unpredictable
> value, the random number should be a function parameter.
>
> In order to test something it must be made testable. Right now your
> function is not.
>
>
> Le mercredi 15 mars 2017 04:45:22 UTC+1, Doug Ireton a écrit :
>>
>> I'm a new Gopher and I'm working through "Learn Go" by Nathan Youngman
>> and trying to TDD the exercises to learn how to write testable Go code.
>>
>> I have a function to return a random spaceline
>>  from a string array.
>>
>> In Go, how do I test functions which depend on random numbers? And, yes,
>> I know that "math/rand" isn't truly random.
>>
>> Is it as simple as setting a seed right before I run my test, e.g.
>> rand.Seed(1)? Do I set rand.Seed(1) at the top of the _test.go file, or
>> at the beginning of each unit test?
>>
>> Also, am I seeding math.rand correctly in the Init() function? Will
>> seeding it in the Init() function override any seeding I do in my tests?
>>
>> My only other thought is to create an interface somehow to mock
>> rand.Intn(), but this seems like overkill and I don't know enough about
>> interfaces to know if this is inadvisable.
>>
> --
> You received this message because you are subscribed to the Google Groups
> "golang-nuts" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to golang-nuts+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>



-- 
Michael T. Jones
michael.jo...@gmail.com

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[go-nuts] Re: How do I test an func which depends on (pseudo)-random numbers, e.g. rand.Intn()?

[Robert Johnstone]

If you are fixing the seed, you are probably over constraining your test. 
 Even though you are calling rand, there must be some post-conditions that 
the function is supposed to guarantee.  In this case, you probably expect 
all options to be returned with equal probability, so you should call the 
function many times, and calculate those probabilities.  There is a whole 
field for statistical testing.


On Tuesday, 14 March 2017 23:45:22 UTC-4, Doug Ireton wrote:
>
> I'm a new Gopher and I'm working through "Learn Go" by Nathan Youngman 
> and trying to TDD the exercises to learn how to write testable Go code.
>
> I have a function to return a random spaceline 
>  from a string array.
>
> In Go, how do I test functions which depend on random numbers? And, yes, I 
> know that "math/rand" isn't truly random.
>
> Is it as simple as setting a seed right before I run my test, e.g. 
> rand.Seed(1)? Do I set rand.Seed(1) at the top of the _test.go file, or 
> at the beginning of each unit test?
>
> Also, am I seeding math.rand correctly in the Init() function? Will 
> seeding it in the Init() function override any seeding I do in my tests?
>
> My only other thought is to create an interface somehow to mock 
> rand.Intn(), but this seems like overkill and I don't know enough about 
> interfaces to know if this is inadvisable.
>

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[go-nuts] Re: How do I test an func which depends on (pseudo)-random numbers, e.g. rand.Intn()?

[mhhcbon]

Hi,

if you wish not declare a struct and its interface, you might simply make a 
function type.

See
https://play.golang.org/p/xuycrPc8sF

Then in your main program, you consume the func type, and inject a func 
impl.

It is not as versatile as interface, but that can make the job.

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[go-nuts] Re: How do I test an func which depends on (pseudo)-random numbers, e.g. rand.Intn()?

[xiiophen]

>
>
> Also, am I seeding math.rand correctly in the Init() function? Will 
> seeding it in the Init() function override any seeding I do in my tests?
>
>
You're using time.Now().UTC().UnixNano()) but

 time.Now().UnixNano() 

or  

time.Now().Unix()

seem just as good

Note UnixNano() may be beyond the accuracy at with the CPU reports - so it 
may well trail zeros .. conversely 1 sec intervals in Unix() may be too 
large .. 

time.Now().UnixNano()/1000 ie microseconds 

or similar could be a good compromise


I assume the reason you are doing this is to get a different set of randoms 
everytime you run - obviously that will make testing next to impossible - 
my suggestion based on experience - test with a set of seeds - ie 1,2,3 in 
rand.Seed  .. testing with just one seed often gave me corner case results 
which misled me on the validity of my code - testing with a set of seeds 
should ensure that your random numbers are "randomly selected"

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[go-nuts] Re: How do I test an func which depends on (pseudo)-random numbers, e.g. rand.Intn()?

[xiiophen]

https://golang.org/pkg/math/rand/#Seed

*"Seed uses the provided seed value to initialize the default Source to a 
deterministic state. If Seed is not called, the generator behaves as if 
seeded by Seed(1). "* 


So even without calling rand.Seed your output will be the same everytime 
you run the program unless you take action to make that different. 

( note that the second value obtained from Rand.IntN is not the same as 
that for the first value obtained from Rand.IntN seeded with the value 2 )

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[go-nuts] Re: How do I test an func which depends on (pseudo)-random numbers, e.g. rand.Intn()?

[Egon]

On Wednesday, 15 March 2017 05:45:22 UTC+2, Doug Ireton wrote:
>
> I'm a new Gopher and I'm working through "Learn Go" by Nathan Youngman 
> and trying to TDD the exercises to learn how to write testable Go code.
>
> I have a function to return a random spaceline 
>  from a string array.
>
> In Go, how do I test functions which depend on random numbers? And, yes, I 
> know that "math/rand" isn't truly random.
>
> Is it as simple as setting a seed right before I run my test, e.g. 
> rand.Seed(1)? Do I set rand.Seed(1) at the top of the _test.go file, or 
> at the beginning of each unit test?
>
> Also, am I seeding math.rand correctly in the Init() function? Will 
> seeding it in the Init() function override any seeding I do in my tests?
>
> My only other thought is to create an interface somehow to mock 
> rand.Intn(), but this seems like overkill and I don't know enough about 
> interfaces to know if this is inadvisable.
>

Write tests for the intended behavior. It seems you are over-specifying the 
test. If something needs to return something random, then test whether the 
result is random enough.

+ Egon 

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[go-nuts] Re: How do I test an func which depends on (pseudo)-random numbers, e.g. rand.Intn()?

[Miki Tebeka]

IMO the right approach is as you suggested - seed once in init.

init is called before your tests, so the seed you do in your test will 
override it.

On Wednesday, March 15, 2017 at 5:45:22 AM UTC+2, Doug Ireton wrote:
>
> I'm a new Gopher and I'm working through "Learn Go" by Nathan Youngman 
> and trying to TDD the exercises to learn how to write testable Go code.
>
> I have a function to return a random spaceline 
>  from a string array.
>
> In Go, how do I test functions which depend on random numbers? And, yes, I 
> know that "math/rand" isn't truly random.
>
> Is it as simple as setting a seed right before I run my test, e.g. 
> rand.Seed(1)? Do I set rand.Seed(1) at the top of the _test.go file, or 
> at the beginning of each unit test?
>
> Also, am I seeding math.rand correctly in the Init() function? Will 
> seeding it in the Init() function override any seeding I do in my tests?
>
> My only other thought is to create an interface somehow to mock 
> rand.Intn(), but this seems like overkill and I don't know enough about 
> interfaces to know if this is inadvisable.
>

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.