exercism bowling challenge
Hi Richard,
I want to reply to some of your insights here, because I think it is
sometimes a matter of personal preference on a problem solution, so I
want to compete with your opinion here. I might be totally wrong, when I
see your way of solution and tell just 'you were right, your solution is
more elegant'. For me it is an approach discussion.
čt 24. 9. 2020 v 15:16 odesílatel Richard O'Keefe raoknz@gmail.com napsal:
Since the problem can be solved, and has been solved, in programming
languages that do not support object-oriented programming, there is
obviously no unique "right" factoring of this problem into classes. I'll
be honest with you: this is the only Pharo exercism I have not tackled, and
the reason is that I just found the problem specification too ugly to care
about. I had better get over that.
We are using OO language to isolate problems into particular objects and
their responsibility. I would say there is no "ugly problem to care about".
Sometimes it can be the "right" factoring of a problem into classes and
sometimes it is too hard and that is the point. I think this exercise is
about finding the right place to delegate problems in fitting classes
that are small enough. Of course, it can be resolved by one "long method"
with a total of less lines of code. But in general I like more verbosity
over one compact algorithm that wouldn't tell readers too much. This is a
personal preference.
The two primary difficulties are
(a) the game is described one ROLL at a time, but scored one FRAME at a
time.
(b) the rolls are recorded from past to future, but the score can only be
calculated from future to past.
Way of evaluation is definitely relevant. Key thing for me was to realize
"I always need to evaluate the current state of things", the same way, when
you play bowling.
- Each frame should be able to tell, if it is completed from the point of
actual throw. So, why not just pass throws by game into the current frame,
that would tell, if it is completed or not? - Frame can calculate score immediately, but bonuses can be later. Maybe it
isn't 'the throw thing' to care about. It is a matter of the game, that can
tell when is the right moment to evaluate bonuses. Game can delegate such
information to the passed frames at the right moment (maybe after each
thow, or maybe when the current frame is completed?). But maybe my previous
sentence can be interpreted as: "this is a big hammer for this solution
already".
The fact that there may be one extra roll at the end which is not
technically
part of any of the 10 frames is just icing on the cake.
But this is algorithmic trickiness, not world model trickiness.
Let's imagine a Frame class.
- You get the first roll of the frame. Fewer than 10 pins are knocked
down.
You cannot complete initialising the Frame yet.
Well, why not just initialize frame by "collection of 1-2 (maybe 3)
throws"? Depending on the actual state of collection, frame just tells the
game "I have enough throws, so you can pass throws to my neighbor". Rules
for completeness can be different, based on circumstances. For me, it is a
matter of frame responsibility to deal with this.
- All 10 pins are knocked down. You cannot determine the score of the
frame yet. This casts some doubt on the idea of the score being part
of the state of the frame. The score actually depends on the next two
THROWS (rolls), not the next frame or the next two frames. This casts
much doubt on the idea of the concept "frame" being useful for the
analysis. At the very least, you will need to manipulate BOTH frames
AND throws (rolls).
When bowling in real, you can determine the score of the current frame. But
bonuses are computed later, so maybe the game should tell (after each
throw), when to evaluate bonus. "Hey, my passed frames, I'm passing you
this information, so evaluate your bonus now".
It looks as though a Frame class may just make things harder.
Since the only thing there is to know about a Throw is how many pins
were knocked down, it doesn't look as though a Throw class is much use
either. This leaves us with a BowlingGame class that just keeps tracks
of rolls and then runs a moderately complex algorithm when it is asked
for the score.
Long past my bed-time or I would get stuck into it.
It depends on how "moderately complex algorithm" will be readable for end
users. Sometimes it is a real struggle to determine "what a hell is going
on here", because the given method just does too many things in one class.
For me, separating throws into frames makes the whole thing more
understandable to readers, because they can imagine something from real
life (real game). Again, this is personal preference. I've seen few times
that "moderately complex algorithms" were extended by other developers into
"severe too-complex algorithms". Problem is that developers don't do
refactoring at the right time (most probably because of lack of time and
time pressure on projects). On the other hand, I understand it isn't a
problem of this artificial exercise and agree that focus should be just
on existing problem statement and not hypothetical problem.
I did this exercise some time ago and when I look at it now, honestly I'm
not satisfied with my solution, especially the way bonuses are stored and
evaluated in normal vs last frame. But here is the thing: it is isolated on
one given place and used on one other place, I can rewrite it maybe more
easily.
On Thu, 24 Sep 2020 at 23:53, Roelof Wobben via Pharo-users <
pharo-users@lists.pharo.org> wrote:
Op 24-9-2020 om 13:42 schreef DavidBajger:
Hi Roelof,
I always wonder, what kind of answer you expect from your prior
statement.
To your question: "Can this plan be working or is there improvements to
this
plan." I can have this answer: Yes, it could be both: working or fail,
but
you don't know before you try.
This exercise is a bit tricky:
- I can recommend to use also LastFrame class, which has specific
handling
of bonuses in last round of game (subclass of Frame). Bowling game can
be
then initialized with array 9 Frame instances and last instance could be
LastFrame. I used specific test methods on frame classes like:
#isFrameComplete, #isLastFrame, #isSpare, #isStrike, #isOpen.
2) Beware that Bowling game should know only necessary things and
delegate
responsibility to its frames. Game itself knows that only if all frames
are
completed, game ends.
3) Total score is sum of all throws+bonuses of individual frames, etc.
Does it help to start with exercise?
David
David Bajger
Sent from: http://forum.world.st/Pharo-Smalltalk-Users-f1310670.html
Thanks,
Maybe I have to make it more clear.
What I trying to ask and what you have answered. Do I use the right
classes or too much classes there.
So if my idea of using these classes are right.
What you wrote is what I had in mind with 2 classes but the idea of a
lastFrame could also be working.
and I also agree with you about the responsibilities of the classes.
Roelof
It turns out that this problem has appeared elsewhere,
at least twice to my knowledge, and I had already solved
the version at Programming Praxis over a year ago.
I actually wrote two solutions, in different programming
languages. One of them was 8 lines. That was the longer.
The Haskell version was just 3 lines.
The Exercism tasks are what they are. They DO provide
drill in using the more conventional aspects of Smalltalk.
They DON'T provide good drill in object-oriented DESIGN.
This particular problem can be solved clearly and cleanly
with a ByteArray and a handful of integers.
The fact that the test harness enforces the raising of
certain exceptions which are nowhere mentioned in the
specification complicates things rather. (As does the
fact that the interface enforced by the test harness
contradicts the interface required by the specification.)
Remember, given a list of rolls, Haskell can do the
whole calculation in three straightforward lines. Getting
the code to work is tricky, not because the calculations
are intrinsically hard, but because the specification is
very confusingly worded.
On Fri, 25 Sep 2020 at 19:23, David Bajger bajger@gmail.com wrote:
Hi Richard,
I want to reply to some of your insights here, because I think it is
sometimes a matter of personal preference on a problem solution, so I
want to compete with your opinion here. I might be totally wrong, when I
see your way of solution and tell just 'you were right, your solution is
more elegant'. For me it is an approach discussion.
čt 24. 9. 2020 v 15:16 odesílatel Richard O'Keefe raoknz@gmail.com
napsal:
Since the problem can be solved, and has been solved, in programming
languages that do not support object-oriented programming, there is
obviously no unique "right" factoring of this problem into classes. I'll
be honest with you: this is the only Pharo exercism I have not tackled, and
the reason is that I just found the problem specification too ugly to care
about. I had better get over that.
We are using OO language to isolate problems into particular objects and
their responsibility. I would say there is no "ugly problem to care about".
Sometimes it can be the "right" factoring of a problem into classes and
sometimes it is too hard and that is the point. I think this exercise is
about finding the right place to delegate problems in fitting classes
that are small enough. Of course, it can be resolved by one "long method"
with a total of less lines of code. But in general I like more verbosity
over one compact algorithm that wouldn't tell readers too much. This is a
personal preference.
The two primary difficulties are
(a) the game is described one ROLL at a time, but scored one FRAME at a
time.
(b) the rolls are recorded from past to future, but the score can only be
calculated from future to past.
Way of evaluation is definitely relevant. Key thing for me was to realize
"I always need to evaluate the current state of things", the same way, when
you play bowling.
- Each frame should be able to tell, if it is completed from the point of
actual throw. So, why not just pass throws by game into the current frame,
that would tell, if it is completed or not? - Frame can calculate score immediately, but bonuses can be later. Maybe
it isn't 'the throw thing' to care about. It is a matter of the game, that
can tell when is the right moment to evaluate bonuses. Game can delegate
such information to the passed frames at the right moment (maybe after each
thow, or maybe when the current frame is completed?). But maybe my previous
sentence can be interpreted as: "this is a big hammer for this solution
already".
The fact that there may be one extra roll at the end which is not
technically
part of any of the 10 frames is just icing on the cake.
But this is algorithmic trickiness, not world model trickiness.
Let's imagine a Frame class.
- You get the first roll of the frame. Fewer than 10 pins are knocked
down.
You cannot complete initialising the Frame yet.
Well, why not just initialize frame by "collection of 1-2 (maybe 3)
throws"? Depending on the actual state of collection, frame just tells the
game "I have enough throws, so you can pass throws to my neighbor". Rules
for completeness can be different, based on circumstances. For me, it is a
matter of frame responsibility to deal with this.
- All 10 pins are knocked down. You cannot determine the score of the
frame yet. This casts some doubt on the idea of the score being part
of the state of the frame. The score actually depends on the next two
THROWS (rolls), not the next frame or the next two frames. This casts
much doubt on the idea of the concept "frame" being useful for the
analysis. At the very least, you will need to manipulate BOTH frames
AND throws (rolls).
When bowling in real, you can determine the score of the current frame.
But bonuses are computed later, so maybe the game should tell (after each
throw), when to evaluate bonus. "Hey, my passed frames, I'm passing you
this information, so evaluate your bonus now".
It looks as though a Frame class may just make things harder.
Since the only thing there is to know about a Throw is how many pins
were knocked down, it doesn't look as though a Throw class is much use
either. This leaves us with a BowlingGame class that just keeps tracks
of rolls and then runs a moderately complex algorithm when it is asked
for the score.
Long past my bed-time or I would get stuck into it.
It depends on how "moderately complex algorithm" will be readable for end
users. Sometimes it is a real struggle to determine "what a hell is going
on here", because the given method just does too many things in one class.
For me, separating throws into frames makes the whole thing more
understandable to readers, because they can imagine something from real
life (real game). Again, this is personal preference. I've seen few times
that "moderately complex algorithms" were extended by other developers into
"severe too-complex algorithms". Problem is that developers don't do
refactoring at the right time (most probably because of lack of time and
time pressure on projects). On the other hand, I understand it isn't a
problem of this artificial exercise and agree that focus should be just
on existing problem statement and not hypothetical problem.
I did this exercise some time ago and when I look at it now, honestly I'm
not satisfied with my solution, especially the way bonuses are stored and
evaluated in normal vs last frame. But here is the thing: it is isolated on
one given place and used on one other place, I can rewrite it maybe more
easily.
On Thu, 24 Sep 2020 at 23:53, Roelof Wobben via Pharo-users <
pharo-users@lists.pharo.org> wrote:
Op 24-9-2020 om 13:42 schreef DavidBajger:
Hi Roelof,
I always wonder, what kind of answer you expect from your prior
statement.
To your question: "Can this plan be working or is there improvements
to this
plan." I can have this answer: Yes, it could be both: working or fail,
but
you don't know before you try.
This exercise is a bit tricky:
- I can recommend to use also LastFrame class, which has specific
handling
of bonuses in last round of game (subclass of Frame). Bowling game can
be
then initialized with array 9 Frame instances and last instance could
be
LastFrame. I used specific test methods on frame classes like:
#isFrameComplete, #isLastFrame, #isSpare, #isStrike, #isOpen.
2) Beware that Bowling game should know only necessary things and
delegate
responsibility to its frames. Game itself knows that only if all
frames are
completed, game ends.
3) Total score is sum of all throws+bonuses of individual frames, etc.
Does it help to start with exercise?
David
David Bajger
Sent from: http://forum.world.st/Pharo-Smalltalk-Users-f1310670.html
Thanks,
Maybe I have to make it more clear.
What I trying to ask and what you have answered. Do I use the right
classes or too much classes there.
So if my idea of using these classes are right.
What you wrote is what I had in mind with 2 classes but the idea of a
lastFrame could also be working.
and I also agree with you about the responsibilities of the classes.
Roelof
Hello,
Maybe developing is not my cup of tea,
I get more and more confused with this discussion.
Roelof
Op 25-9-2020 om 14:37 schreef Richard O'Keefe:
It turns out that this problem has appeared elsewhere,
at least twice to my knowledge, and I had already solved
the version at Programming Praxis over a year ago.
I actually wrote two solutions, in different programming
languages. One of them was 8 lines. That was the longer.
The Haskell version was just 3 lines.
The Exercism tasks are what they are. They DO provide
drill in using the more conventional aspects of Smalltalk.
They DON'T provide good drill in object-oriented DESIGN.
This particular problem can be solved clearly and cleanly
with a ByteArray and a handful of integers.
The fact that the test harness enforces the raising of
certain exceptions which are nowhere mentioned in the
specification complicates things rather. (As does the
fact that the interface enforced by the test harness
contradicts the interface required by the specification.)
Remember, given a list of rolls, Haskell can do the
whole calculation in three straightforward lines. Getting
the code to work is tricky, not because the calculations
are intrinsically hard, but because the specification is
very confusingly worded.
On Fri, 25 Sep 2020 at 19:23, David Bajger <bajger@gmail.com> wrote:
Hi Richard,
I want to reply to some of your insights here, because I think it is sometimes a matter of personal preference on a problem solution, so I want to compete with your opinion here. I might be totally wrong, when I see your way of solution and tell just 'you were right, your solution is more elegant'. For me it is an approach discussion.
čt 24. 9. 2020 v 15:16 odesílatel Richard O'Keefe <raoknz@gmail.com> napsal:
Since the problem can be solved, and has been solved, in programming languages that do not support object-oriented programming, there is obviously no unique "right" factoring of this problem into classes. I'll be honest with you: this is the only Pharo exercism I have not tackled, and the reason is that I just found the problem specification too ugly to care about. I had better get over that.
We are using OO language to isolate problems into particular objects and their responsibility. I would say there is no "ugly problem to care about". Sometimes it can be the "right" factoring of a problem into classes and sometimes it is too hard and that is the point. I think this exercise is about finding the right place to delegate problems in fitting classes that are small enough. Of course, it can be resolved by one "long method" with a total of less lines of code. But in general I like more verbosity over one compact algorithm that wouldn't tell readers too much. This is a personal preference.
The two primary difficulties are
(a) the game is described one ROLL at a time, but scored one FRAME at a time.
(b) the rolls are recorded from past to future, but the score can only be
calculated from future to past.
Way of evaluation is definitely relevant. Key thing for me was to realize "I always need to evaluate the current state of things", the same way, when you play bowling.
Each frame should be able to tell, if it is completed from the point of actual throw. So, why not just pass throws by game into the current frame, that would tell, if it is completed or not?
Frame can calculate score immediately, but bonuses can be later. Maybe it isn't 'the throw thing' to care about. It is a matter of the game, that can tell when is the right moment to evaluate bonuses. Game can delegate such information to the passed frames at the right moment (maybe after each thow, or maybe when the current frame is completed?). But maybe my previous sentence can be interpreted as: "this is a big hammer for this solution already".
The fact that there may be one extra roll at the end which is not technically
part of any of the 10 frames is just icing on the cake.
But this is algorithmic trickiness, not world model trickiness.
Let's imagine a Frame class.
- You get the first roll of the frame. Fewer than 10 pins are knocked down.
You cannot complete initialising the Frame yet.
Well, why not just initialize frame by "collection of 1-2 (maybe 3) throws"? Depending on the actual state of collection, frame just tells the game "I have enough throws, so you can pass throws to my neighbor". Rules for completeness can be different, based on circumstances. For me, it is a matter of frame responsibility to deal with this.
- All 10 pins are knocked down. You cannot determine the score of the
frame yet. This casts some doubt on the idea of the score being part
of the state of the frame. The score actually depends on the next two
THROWS (rolls), not the next frame or the next two frames. This casts
much doubt on the idea of the concept "frame" being useful for the
analysis. At the very least, you will need to manipulate BOTH frames
AND throws (rolls).
When bowling in real, you can determine the score of the current frame. But bonuses are computed later, so maybe the game should tell (after each throw), when to evaluate bonus. "Hey, my passed frames, I'm passing you this information, so evaluate your bonus now".
It looks as though a Frame class may just make things harder.
Since the only thing there is to know about a Throw is how many pins
were knocked down, it doesn't look as though a Throw class is much use
either. This leaves us with a BowlingGame class that just keeps tracks
of rolls and then runs a moderately complex algorithm when it is asked
for the score.
Long past my bed-time or I would get stuck into it.
It depends on how "moderately complex algorithm" will be readable for end users. Sometimes it is a real struggle to determine "what a hell is going on here", because the given method just does too many things in one class. For me, separating throws into frames makes the whole thing more understandable to readers, because they can imagine something from real life (real game). Again, this is personal preference. I've seen few times that "moderately complex algorithms" were extended by other developers into "severe too-complex algorithms". Problem is that developers don't do refactoring at the right time (most probably because of lack of time and time pressure on projects). On the other hand, I understand it isn't a problem of this artificial exercise and agree that focus should be just on existing problem statement and not hypothetical problem.
I did this exercise some time ago and when I look at it now, honestly I'm not satisfied with my solution, especially the way bonuses are stored and evaluated in normal vs last frame. But here is the thing: it is isolated on one given place and used on one other place, I can rewrite it maybe more easily.
On Thu, 24 Sep 2020 at 23:53, Roelof Wobben via Pharo-users <pharo-users@lists.pharo.org> wrote:
Op 24-9-2020 om 13:42 schreef DavidBajger:
> Hi Roelof,
> I always wonder, what kind of answer you expect from your prior statement.
> To your question: "Can this plan be working or is there improvements to this
> plan." I can have this answer: Yes, it could be both: working or fail, but
> you don't know before you try.
>
> This exercise is a bit tricky:
> 1) I can recommend to use also LastFrame class, which has specific handling
> of bonuses in last round of game (subclass of Frame). Bowling game can be
> then initialized with array 9 Frame instances and last instance could be
> LastFrame. I used specific test methods on frame classes like:
> #isFrameComplete, #isLastFrame, #isSpare, #isStrike, #isOpen.
> 2) Beware that Bowling game should know only necessary things and delegate
> responsibility to its frames. Game itself knows that only if all frames are
> completed, game ends.
> 3) Total score is sum of all throws+bonuses of individual frames, etc.
>
> Does it help to start with exercise?
> David
>
>
>
>
>
> -----
> David Bajger
> --
> Sent from: http://forum.world.st/Pharo-Smalltalk-Users-f1310670.htmlThanks,
Maybe I have to make it more clear.
What I trying to ask and what you have answered. Do I use the right
classes or too much classes there.
So if my idea of using these classes are right.What you wrote is what I had in mind with 2 classes but the idea of a
lastFrame could also be working.
and I also agree with you about the responsibilities of the classes.Roelof
Thanks for your reply! Anyway.. wow, 3 lines of code, I'd like to see that! I
didn't really get into FLP (except some small howeworks on University and
that's alraeady forgotten), so I don't know, what is the advantage here. I
thought that Pharo (Smalltalk) can be considered as functional language too,
because of blocks. If there wouldn't be any exception handling, would Pharo
code be so compact as well (e.g. by using #inject:into:)?
David Bajger
Sent from: http://forum.world.st/Pharo-Smalltalk-Users-f1310670.html
Yes, my thoughts - since I'm not as technical as the rest of you:
BowlingGame class
"runs the game a calls each frame to 'process'"
- frames (1-10 aFrame objects)
- score (aScore object)
- currentFrame
Frame class
"frame handles #throws, pins remaining, and lets Game know when to advance
to next frame - frameEnd"
- frame#
- throws (1-3 aThrow objects - aFrame controls 10th frame 3rd throw)
- score (numeric)
Throw class
- frame (the frame I belong to)
- throw# (incremental numeric 1-3)
- pinsKnockedDown (0-10)
Score class
- frames (same as BowlingGame class)
- methods to calculate score, once a frame score has been determined, it is
stored on the frame and the score calc simply picks that up and moves
forward on each successive score calc (after each throw - called from
inside of aFrame)
This will be a fun little diversion. I'll create a simple UI and engine
that will randomly return a number of pinsKnockedDown when pressing the
spacebar to roll the next ball.
On Fri, Sep 25, 2020 at 3:11 PM DavidBajger bajger@gmail.com wrote:
Thanks for your reply! Anyway.. wow, 3 lines of code, I'd like to see
that! I
didn't really get into FLP (except some small howeworks on University and
that's alraeady forgotten), so I don't know, what is the advantage here. I
thought that Pharo (Smalltalk) can be considered as functional language
too,
because of blocks. If there wouldn't be any exception handling, would Pharo
code be so compact as well (e.g. by using #inject:into:)?
David Bajger
Sent from: http://forum.world.st/Pharo-Smalltalk-Users-f1310670.html
--
Russ Whaley
whaley.russ@gmail.com