Finished 17/01/2021

Clean Code

Robert C. Martin

⭐⭐

• IT

Addendum: https://github.com/zedr/clean-code-python#dependency-inversion-principle-dip

You do not write clean code by following a simple set of rules. You will not become “craftsman developers” by learning a simple list of heuristics. Professionalism and skill come from the values that guide the discipline.

Dense read, with plenty of Java code I did not fully follow, chapters unfamiliar to me, yet solid gems for programming. I will leave the final extract that sums up the advice; overall it seemed a very useful book for a programmer.

Comments

• No inappropriate info: comments are only for technical explanations of the code
• Remove obsolete comments; they only confuse
• Remove redundant comments; they distract if the code is well written
• Do not keep commented-out code; people leave it in thinking it might be useful. Avoid it.

Environment (understood little)

• To build, a single command should suffice; you should not have to gather extra files
• Testing should be simple; start it with a single click

Functions

• THEY MUST DO ONE THING ONLY
• Do not have too many arguments, max 3
• No output arguments (Java-specific); prefer input
• No booleans as arguments; that often means it does more than one thing
• No dead functions (never called); they pollute the code

General

• Avoid multiple languages in the code even if you can; it creates confusion (like HTML in comments)
• Avoid surprises for a reader: functions should do what you expect
• Do not rely on intuition; set boundaries and explore edge cases
• Do not disable safeguards and warnings; it may work now but causes problems later
• DUPLICATION: REMOVE DUPLICATES; THEY CAN BECOME FUNCTIONS/CLASSES
• Set up abstractions well (Java?)
• Excess information: hide service data, variables, and classes the user should not see if they are background details. They do not need to sit in main if they are not meaningful
• Bury dead code (e.g., except blocks that never run or unnecessary if cases)
• Vertical separation: declare variables just above their first use so you do not have to search
• Inconsistency: adopt the same structure everywhere; be predictable
• Artificial coupling: do not put constants/functions where they do not belong
• Obscured intent: it might seem “cool” to use Hungarian notation or skip spaces, but be understandable
• Misplaced responsibilities: things go where you expect them. const pi will live in a Math class
• Use descriptive variables: names should say what the function or variable does, even if long
• As above, watch for ambiguity: what does addDate do? better use addDaysToDate
• Use self-explanatory constants: in ambiguous cases, assign the number to a new descriptive constant
• Encapsulate conditional constructs: for long if-and-or chains, use a function that returns true or false
• Avoid negatives in conditional constructs: for convenience, do not use negations
• Keep configurable data at higher levels: if a constant might change, keep it in main
• Avoid transitive navigation: modules should not be too intertwined; limit communication among them. Ideally have a central module that queries the others independently

Java (understood little)

• If you import a lot, use the wildcard (from datetime import *)
• Do not inherit constants: they should not come from some function somewhere; declare them in main

Names

• CHOOSE DESCRIPTIVE, UNAMBIGUOUS NAMES
• Use standard naming where possible
• Long names are fine, but do not overdo it or you risk confusion
• Avoid encodings like _f or _v (Java)

Tests

• A test package should contain everything needed to test the program
• Use a coverage tool; it shows which lines are exercised by tests
• Do not skip tests you consider trivial
• An ignored test signals ambiguity
• Consider when problems seem to suggest a pattern
• Tests should be fast