Discuss selective application of representer

[mikedamay]

Whilst many representer rewrites need to be applied to all exercises there are cases where a rewrite should be disabled for a particular exercise. For example, the readonly modifier should not be removed from code in the constants exercise where the concept is introduced.

The requirement to disable a rewrite for the exercise in which the related concept is introduced is fairly clear cut. However, the related issue of whether a rewrite should be disabled where its inclusion would prevent good/bad practice being reported to the student needs some discussion.

Which of the following is Excercism's position:

A. "Once a concept (which may be associated with some good practice or other) has been introduced then a rewrite that would suppress the good practice should be disabled and bad practice reported to the student"

or

B. "Reports to the student should be concerned only with the concepts that the exercise is designed to demonstrate"

Good practice relates to the writing of robust, maintainable and performant code.

These are the current C# rewrites:

Rewrite	Concept	Excercise	Good Practice Importance
Add Braces	None	None	None
Expression extensions	??	??	??
Invert Neg Conditions	None	None	None
exponnt form to ucase	None	None	None
normalize identifiers	basics	basics	High
normalize whitespace	None	None	None
remove comments	None	None	None
optional parentheses	None	None	None
remove readonly	readonly	constants	High
remove usings	accessibility	namespaces	None
simplify bool equals	booleans	booleans	Medium
String->string	strings	strings	Low
Fully Qualified Name	basics	basics	Very Low

Although Identifiers are a major element in good practice and rewriting will eface them, the practicalities of representations rule out dealing with them as we would wish in the concept exercises.

Of the current rewrites, removing readonly has a fairly high impact on good practice and simplifying boolean expressions (if (x == true) -> if (x)) is medium to high. casing string less important and reducing fully qualified names is low (maybe because it would probably take care of itself - I think we are entitled to rely on that).

Three questions:

1. Should we disable rewrites for exercises where there is a dependency on a concept with high or medium importance?

2. Should we increase the number of exercises where concepts which have a high or medium good practice importance are taken as a dependency?

3. What kind of rewrites will be added to the representer in the future?

[ErikSchierboom]

Which of the following is Excercism's position:

A. "Once a concept (which may be associated with some good practice or other) has been introduced then a rewrite that would suppress the good practice should be disabled and bad practice reported to the student"

or

B. "Reports to the student should be concerned only with the concepts that the exercise is designed to demonstrate"

This is an excellent question, and something that we've discussed a bit when analyzers were being developed. Different tracks have built their analyzers differently, with some tracks going all-in on posting as many comments as can be gathered, whereas others have more selectively commented on things. Some did a tiered approach, where they would first only comment on important issues, and then if they were fixed, they would comment on the less important issues (which is how the C# analyzer worked). We can go either way with this. @iHiD your thoughts?

The requirement to disable a rewrite for the exercise in which the related concept is introduced is fairly clear cut. However, the related issue of whether a rewrite should be disabled where its inclusion would prevent good/bad practice being reported to the student needs some discussion.

The way I'm thinking about this is that the above rewrites are all non-essential to most exercises (as you mentioned). We should indeed selectively disables rewrites if they remove what we want to comment on. However, for all the other exercises, I feel like we could have the representer use the rewrites. In my mind, commenting on things like invalid white space, optional parentheses, missing readonly modifiers and such is the job of the C# analyzer, not of the representer. My reasoning here is that the aforementioned things are things to comment on for all exercises, not just a specific representation. This way, mentors that comment on representations only have to deal with what the exercise is actually about, and not bother with things that are applicable to any exercise. Note that we would still have to build in something in the analyzer to not comment on concept that the student doesn't yet know about. We may be able to use the concept tree as defined in the config.json for that.

Does the above make sense?

Although Identifiers are a major element in good practice and rewriting will eface them, the practicalities of representations rule out dealing with them as we would wish in the concept exercises.

This is true. Traditionally most mentors have commented on naming of identifiers, which while important, isn't necessarily about becoming fluent in a language.

What kind of rewrites will be added to the representer in the future?

This is hard to gauge, as it will probably be best informed by seeing how representations will actually vary. Things that I can think of are:

• Converting expression-bodied members to regular, block-based members
• Ordering the classes/members in a predetermined way

[iHiD]

@iHiD your thoughts?

This needs more thought that I'm going to have before my break to be honest, as my brain is too overloaded to consider all the implications.

I think though that there may be different rules for Concept and Practice exercises (e.g. max one comment for CEs, multiple for PEs; CEs comments focus on using that concept only, PEs highlight bad things).

The key consideration would be that CEs should generally all have optimal solutions, whereas PEs generally have multiple interesting possible approaches.