884 grudger - with 1 to 3 issues #993

eric-s-s · 2017-04-28T06:24:42Z

grudger has one issue and possibly three.

ForgetfulGrudger: changed in PR
- docstring and code did not match. docstring said grudge for 10 rounds and code said grudge for 11 rounds
- changed code to fit docstring
GeneralSoftGrudger: changed in PR
- possibly an issue IF edge cases are valid.
- I tested for edge cases of n=0, d=0 and c=0. original code did not follow expected. updated code.
Possible issue not changed:
- should ForgetfulGrudger and SoftGrudger inherit from GeneralSoftGrudger? they are GeneralSoftGrudger(n=1, d=10, c=0) and GeneralSoftGrudger(n=1, d=4, c=2)

eric-s-s · 2017-04-28T06:25:12Z

axelrod/strategies/grudger.py

-    only punishes after its opponent defects a specified amount of times consecutively.
-    The punishment is in the form of a series of defections followed by a 'penance' of
-    a series of consecutive cooperations.
+    A generalization of the SoftGrudger strategy. SoftGrudger punishes by


changed docstring to fit line-llen = 80.

eric-s-s · 2017-04-28T06:26:27Z

axelrod/strategies/grudger.py

@@ -63,16 +63,16 @@ def __init__(self) -> None:
    def strategy(self, opponent: Player) -> Action:
        """Begins by playing C, then plays D for mem_length rounds if the
        opponent ever plays D."""
-        if self.grudge_memory >= self.mem_length:


strategy docstring says defect for 10 rounds. strategy code said detect for 11 rounds. I changed the code to fit the docstring.

eric-s-s · 2017-04-28T06:30:30Z

axelrod/strategies/grudger.py

@@ -293,16 +294,18 @@ def strategy(self, opponent: Player) -> Action:
        The punishment is in the form of 'd' defections followed by a penance of
        'c' consecutive cooperations.
        """


This startegy was changed to fit the edge cases where n=0 or d=0 or c=0. (see the tests at link for an example line 309 of tests). If the edge cases are invalid, then this change is totally unnecessary.

If the edge cases are invalid,

I don't understand what you mean.

sorry. that wasn't clear.

are n=0, d=0 and c=0 legal inputs?

if they are, the original code failed on n=0 and d=0. This change is to fix that.

if they are not allowed as inputs, then the original code works just fine and i can remove this change. (but perhaps it would be useful to add to n>=1 and d>=1 to the docstring)

and on reflection, those cases would never be called. i will remove the changes.

eric-s-s · 2017-04-28T06:43:26Z

axelrod/tests/strategies/test_grudger.py

-
+        # the default follows SoftGrudger
+        super(TestGeneralSoftGrudger, self).test_strategy()
+


@drvinceknight - in the previous PR you asked me why i changed from the original test_strategy to the following three tests. so I thought it over and here is my reasoning.

1 - I thought tests physically separated from each other and the default behavior tests were more clear.
2 - for n != 1 I wanted tests that showed the edges at and below the threshold. (and what happens with constant defections, which, for reasons unknown to me, I kept getting surprised about)

whether or not these are good reasons, I do think it useful to have this inherit from TestSoftGrudger which will test the default behavior and reset

I am a bit weary of PRs that change things that have been (very recently) added when the motivation for change seems stylistic. This takes a significant cognitive effort by the reviewers to make sure we're not breaking anything. If you recall there were a fair few times I had to explain why something was done a particular way in the fingerprint class.

Please keep these as they are. #884 is a simple issue: refactor tests to use versus_test. If you catch a case that is missing by all means add it.

whether or not these are good reasons, I do think it useful to have this inherit from TestSoftGrudger which will test the default behavior and reset

test_reset can actually be removed entirely by the way, this is now being tested by the parent class TestPlayer.

i apologize. i have great difficulty figuring these kinds of things out. this is an entirely new set of social rules far outside of my usual experience. the very explicit "this is annoying for this reason and causes this kind of trouble" is quite helpful. thank you for that.

i'll wait on replies to the other two questions and revert to the original in the next PR.

👍 Sorry for the late reply, I've been busy on other things. Looking through it now :)

eric-s-s · 2017-04-28T06:44:22Z

axelrod/tests/strategies/test_grudger.py

+        actions = actions_start + subsequent * 5
+        self.versus_test(opponent, expected_actions=actions,
+                         init_kwargs=init_kwargs)
+


Are these edge cases valid or beyond the scope of what this class should be allowed to do?

What do you meant by "valid"?

are n=0, d=0 and c=0 legal/allowed inputs? or is it required that n>=1, d>=1, c>=1?

eric-s-s · 2017-04-28T06:46:01Z

axelrod/tests/strategies/test_grudger.py

+        actions = actions_start + subsequent * 5
+        self.versus_test(opponent, expected_actions=actions,
+                         init_kwargs=init_kwargs)
+


I noticed in GoByMajority that it change self.name to reflect the parameters but that this updates repr to reflect the parameters. is one preferable to the other?

Let's not change that for now.

drvinceknight · 2017-04-28T18:21:12Z

axelrod/strategies/grudger.py

@@ -63,16 +63,16 @@ def __init__(self) -> None:
    def strategy(self, opponent: Player) -> Action:
        """Begins by playing C, then plays D for mem_length rounds if the
        opponent ever plays D."""
-        if self.grudge_memory >= self.mem_length:


drvinceknight · 2017-04-28T18:22:21Z

axelrod/strategies/grudger.py

@@ -293,16 +294,18 @@ def strategy(self, opponent: Player) -> Action:
        The punishment is in the form of 'd' defections followed by a penance of
        'c' consecutive cooperations.
        """


If the edge cases are invalid,

I don't understand what you mean.

drvinceknight · 2017-04-28T18:32:00Z

axelrod/tests/strategies/test_grudger.py

+        actions = actions_start + subsequent * 5
+        self.versus_test(opponent, expected_actions=actions,
+                         init_kwargs=init_kwargs)
+


What do you meant by "valid"?

drvinceknight · 2017-04-28T18:32:30Z

axelrod/tests/strategies/test_grudger.py

+        actions = actions_start + subsequent * 5
+        self.versus_test(opponent, expected_actions=actions,
+                         init_kwargs=init_kwargs)
+


Let's not change that for now.

drvinceknight · 2017-04-28T18:38:09Z

axelrod/tests/strategies/test_grudger.py

-
+        # the default follows SoftGrudger
+        super(TestGeneralSoftGrudger, self).test_strategy()
+


I am a bit weary of PRs that change things that have been (very recently) added when the motivation for change seems stylistic. This takes a significant cognitive effort by the reviewers to make sure we're not breaking anything. If you recall there were a fair few times I had to explain why something was done a particular way in the fingerprint class.

Please keep these as they are. #884 is a simple issue: refactor tests to use versus_test. If you catch a case that is missing by all means add it.

whether or not these are good reasons, I do think it useful to have this inherit from TestSoftGrudger which will test the default behavior and reset

test_reset can actually be removed entirely by the way, this is now being tested by the parent class TestPlayer.

drvinceknight

1 tiny request.

drvinceknight · 2017-04-29T19:24:18Z

axelrod/tests/strategies/test_grudger.py

+        # D ten times and then continue to check for defections.
+        opponent = axl.Cooperator()
+        actions = [(C, C)] * 20
+        self.versus_test(opponent, expected_actions=actions)


These look great but we're losing the attrs test above. Could you add them in?

From: http://axelrod.readthedocs.io/en/latest/tutorials/contributing/strategy/writing_test_for_the_new_strategy.html:

The function versus_test also accepts a dictionary parameter of attributes to check at the end of the match. For example this test checks if the player’s internal variable opponent_class is set to "Cooperative":

Let me know if I can assist.

drvinceknight

This looks great: thanks @eric-s-s 👍

eric-s-s added 2 commits April 27, 2017 23:11

grudger: changed ForgetfulGrudger. tests now use self.versus_test

d2ed293

GeneralSoftGrudger changed to cover edge cases. edge case tests included

9d6ac1d

eric-s-s commented Apr 28, 2017

View reviewed changes

drvinceknight requested changes Apr 28, 2017

View reviewed changes

eric-s-s added 2 commits April 29, 2017 13:23

reverted GeneralSoftGrudger tests

7841213

removed changes for cases that will not occur

63bcaa2

drvinceknight requested changes Apr 29, 2017

View reviewed changes

eric-s-s added 2 commits April 30, 2017 11:53

added attrs to tests in ForgetfulGrudger

2ce541c

closer to original TestForgetfulGrudger

4351ca0

drvinceknight approved these changes Apr 30, 2017

View reviewed changes

drvinceknight added the ready-to-merge label Apr 30, 2017

meatballs merged commit 76ebcaa into Axelrod-Python:master Apr 30, 2017

eric-s-s deleted the 884-grudger branch April 30, 2017 13:44

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

884 grudger - with 1 to 3 issues #993

884 grudger - with 1 to 3 issues #993

eric-s-s commented Apr 28, 2017

eric-s-s Apr 28, 2017

eric-s-s Apr 28, 2017

drvinceknight Apr 28, 2017

eric-s-s Apr 28, 2017 •

edited

drvinceknight Apr 28, 2017

eric-s-s Apr 29, 2017

eric-s-s Apr 29, 2017

eric-s-s Apr 28, 2017

drvinceknight Apr 28, 2017

eric-s-s Apr 29, 2017

eric-s-s Apr 29, 2017 •

edited

drvinceknight Apr 29, 2017

eric-s-s Apr 28, 2017

drvinceknight Apr 28, 2017

eric-s-s Apr 29, 2017

eric-s-s Apr 28, 2017

drvinceknight Apr 28, 2017

drvinceknight Apr 28, 2017

drvinceknight Apr 28, 2017

drvinceknight Apr 28, 2017

drvinceknight Apr 28, 2017

drvinceknight Apr 28, 2017

drvinceknight left a comment

drvinceknight Apr 29, 2017

drvinceknight left a comment


		# the default follows SoftGrudger
		super(TestGeneralSoftGrudger, self).test_strategy()

884 grudger - with 1 to 3 issues #993

884 grudger - with 1 to 3 issues #993

Conversation

eric-s-s commented Apr 28, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

eric-s-s Apr 28, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

eric-s-s Apr 29, 2017 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

drvinceknight left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

drvinceknight left a comment

Choose a reason for hiding this comment

eric-s-s Apr 28, 2017 •

edited

eric-s-s Apr 29, 2017 •

edited