rspec users - Dec 2009 - State-based expectations (as per jMock in GOOS)

Hi all

I''m working my way through Growing Object-Oriented Software[1],
currently at the start of chapter 15.  Chapter 14 introduces a concept I
haven''t seen before, state-based expectations based on sent messages.

The principle appears to be 

  object O has sent message M => O is in state S

followed by

  O is in state S     => (expectation E met) passes spec
  O is not in state S => (expectation E met) violates spec

Has anyone tried this with RSpec (mocks) yet?  Just wondered if it was doable
with the current syntax support or would need an extension.

I thought I''d fire this off now to get ideas, as I won''t be
coding using this technique until I''ve finished the book (a week at
least, I imagine).

Cheers
Ashley

[1] http://www.amazon.co.uk/gp/product/0321503627/


-- 
http://www.patchspace.co.uk/
http://www.linkedin.com/in/ashleymoran

On 18 Dec 2009, at 14:35, Ashley Moran wrote:
> The principle appears to be 
>  object O has sent message M => O is in state S
> followed by
>  O is in state S     => (expectation E met) passes spec
>  O is not in state S => (expectation E met) violates spec
Can you elaborate? From a position of no knowledge, the most obvious question to
me is: why would I care about the state of O? Either the change in O''s
state is observable through its behaviour (in which case I specify that
behaviour) or it''s not (in which case I don''t care).

Cheers,
-Tom

On Dec 18, 2009, at 2:46 pm, Tom Stuart wrote:
> Can you elaborate? From a position of no knowledge, the most obvious
question to me is: why would I care about the state of O? Either the change in
O''s state is observable through its behaviour (in which case I specify
that behaviour) or it''s not (in which case I don''t care).
The example given is for an auction sniper (S).  So, 

  Given S has just notified one of its listeners that it''s bidding
  When S is informed that a bid has been made by a competitor
  And the auction is closed
  Then S should notify its listeners that it lost the auction

  Given S has just notified one of its listeners that it''s bidding
  When S is informed that its bid is highest
  And the auction is closed
  Then S should notify its listeners that it won the auction

The last one looks like the code below in Java[1], but note this is further on
from the code in chapter 14.

Interestingly, writing it as GWT transforms the style of the expectations.  eg,

  "When S is informed that a bid has been made by a competitor"

replaces (something like)

  Allow S to notify its listeners that it''s in the bidding state

Am I making sense?  Not sure if these examples capture the essense.

Ashley


  @Test public void 
  reportsWonIfAuctionClosesWhenWinning() { 
    allowingSniperBidding();
    allowingSniperWinning();
    ignoringAuction();
    
    context.checking(new Expectations() {{ 
      atLeast(1).of(sniperListener).sniperStateChanged(
        new SniperSnapshot(ITEM_ID, 135, 135, WON));
      when(sniperState.is("winning"));
    }}); 
    
    sniper.currentPrice(123, 12, PriceSource.FromOtherBidder);
    sniper.currentPrice(135, 45, PriceSource.FromSniper); 
    sniper.auctionClosed(); 
  }

  private void allowingSniperBidding() {
    allowSniperStateChange(BIDDING, "bidding");
  }
 
  private void allowingSniperWinning() {
    allowSniperStateChange(WINNING, "winning");
  }
 
  private void allowSniperStateChange(final SniperState newState, final String
oldState) {
    context.checking(new Expectations() {{ 
     
allowing(sniperListener).sniperStateChanged(with(aSniperThatIs(newState)));
then(sniperState.is(oldState));
    }});
  }

[1]
http://github.com/sf105/goos-code/blob/master/test/unit/test/auctionsniper/AuctionSniperTest.java

-- 
http://www.patchspace.co.uk/
http://www.linkedin.com/in/ashleymoran

On 18 Dec 2009, at 14:46, Tom Stuart wrote:
> Can you elaborate? From a position of no knowledge, the most obvious
question to me is: why would I care about the state of O? Either the change in
O''s state is observable through its behaviour (in which case I specify
that behaviour) or it''s not (in which case I don''t care).

On further investigation (aka reading up to chapter 24) it''s apparent
that this state-based testing is effectively a flexible form of message
expectation ordering.  What''s clever is that it gives an intermediate
level between no ordering and sequential ordering.  It''s like ordering
"sets" of expectations, at least that''s how it appears to me
at first glance.

Worth more investigation, I think...

Ashley

-- 
http://www.patchspace.co.uk/
http://www.linkedin.com/in/ashleymoran