|
|
|
The one problem I have with the everything-on-one-line approach style of programming (which I also do a lot) is that it's impossible to debug.
You don't get intermediate values and you don't get to set breakpoints at specific functions.
The readability aspect is not that big of a deal, I can read it just fine (usually).
However, since you write code only once and read and debug it a gazillion times I prefer to write my code so I can step through it.
Sometimes I put it back together for readability afterwards.
For example:
var filtered = myList.Where(...);
var ordered = filtered.OrderBy(...);
var result = ordered.ToList();
return result;
return myList
.Where(...)
.OrderBy(...)
.ToList();
|
|
|
|
|
All that depends on the debugger. E.g. in Visual Studio, if you write your code as
string Foo(int x) {
if (x == 1) return "one";
else return "not one";
} and you only want to break when "one" is returned, mark the statement return "one"; with the mouse and hit F9. It will not break for the "if" tests, but only if the true branch is taken.
|
|
|
|
|
Yeah, but I believe that's a new feature since VS2017
Just like you can now debug in-line lambda's.
It wasn't always like that (and still sometimes isn't).
|
|
|
|
|
This I can parse and validate, at a glance, without trying:
string Foo(int x)
{
if(x == 1)
{
return "one";
}
else
{
return "not one";
}
}
}
This requires me to actually read the words on screen and consciously think about them:
string Foo(this int x) => (x == 1) ? "one" : "not one";
It's not "more clear" on a macro level. It's "more clear" to you personally.
If you were one of my juniors, I'd reprimand you for using an esoteric personal style at work.
If you were a senior colleague, I'd just skip all formalities and just stab you with a kitchen knife.
|
|
|
|
|
Do you really have parsing problems with
string Foo(int x) {
if (x == 1) return "one";
else return "not one";
} This style gives me a third as many source "code lines" to relate to. I can easily overview the entire function. In a larger function, a less whitespaced layout makes it much easier to spot the start/end of loops and condition blocks etc.
|
|
|
|
|
I can parse it line by line at a normal reading speed. Takes about 3-4 seconds the first time I see it.
The initial style I can parse and verify without having to read a single word. Takes me about 250 milliseconds.
The benefit of having spacious functions is that mistakes become embarrassingly easy to spot.
Likewise, properly written functions become very easy to parse, because they look so damn familiar.
I find "less lines of code" a dubious benefit.
A cramped style is more time-consuming to debug, both in terms of setting breakpoints and retracing logged exceptions.
Also, I definitely agree functions should not grow too big.. but stacking multiple logical steps on a single line only gives you the illusion of small functions.
I say: be spacious, let functions grow, and flag them for refactoring if they don't fit on a single page.
The point is to write code that's self-evident +6 months in the future.
Quick to read, amend and fix if need be.
|
|
|
|
|
KBZX5000 wrote: because they look so damn familiar. That is the main point here.
"Readability" is not as rationally and objectively defined as we would like to think; it is highly personal. What you learned at school. Universities (at least here in Norway) tend to stress whitespace as the most essential element of good programming, so yuong programmers thend to frown at lines exceeding 30 characters.
When the code is spread over three times as many lines, I spend a lot of attention searching for the match: Hey, there is an "else" here - where was the matching "if"? Not on the line above, not two or three lines above, but four lines above.
if/else is usually fairly simple to handle, though. Braces are far worse. When you unroll seven levels of nested blocks, exactly where is the start of the block closed by the fourth one in the unstacking? You don't know which keyword to look for; everything starts with a brace. If the match is 16 lines higher up, it is much easier to match than if it is 67 lines higher up!
Also, I avoid superflous braces: There is no reason to make a block for every single-statement if, while or for clause. If you try to find the matching opening brace (like above), it is far easier if there are five candidates, as compared to twenty three.
(Exceptions are exceptions: The language "designers" did not learn from Pascal that a statement IS a block; they defined a block concept excluding a statement from being a block. Then came the excptional people declaring the exception mechanism to require blocks for all its parts. I consider statements not being blocks as one of the very big mistakes in the formal definition of the c language.)
It is not just for the braces: I really dislike "the rightward migration of source code", as it was once described. To some people, ideal source code has a left margin like a funnel (and sometimes, the underside of the funnel makes you think that you are programming lisp). Matching opening and closing braces is close to impossible except for the two or three innermost levels.
The fewer indentation levels, the better! Following a "return" statement by an "else" is meaningless. I kept it the code example above, because I know lot of coders would bark if I removed it, even though it really has no meaning, and is in a sense misleading: It seems to suggest that the code following the if(){} else{} is unconditional, regardless of the if-outcome. You actually have to open up that if-clause to discover that not only the statement in the else-clause itself but everything that follow the "else" it is conditional on that if-test.
Very often, funnel code is some variation of if - else if - else if ... but with such complex logic that a simple elseif-sequence won't do. The main point is: if (so-and-so), then do it and you are done! All the rest of the funnel is for other cases! Several times when I am the only one handling some code, in C I set up a "#define once i=1;i==1;i++" so that I can program like:
for (once) {
if (complex condition) {
...do this and that...
break;
if (another complex condition) {
...do something else...
break;
... as many processing alternatives as relevant ...
} The for loop is a dummy: Because C doesn't provide any mechamism for breaking out of an arbitrary block, we must create a dummy loop to allow the break.
Good old Z.200 CHILL has a very nice feature: Any block (including single statements) could be assigned a label; a function name is a label. The EXIT statement breaks out of the block being so labeled. Also, at the end of the block, you can optionally add the label name for easier matching. The compiler processes it, and if you have placed it wrong, the compiler barfs at you. Yet another thing: Blocks are not enclosed in anonymous braces, but with identifying keywords that tells what kind of block ends here. CHILL doesn't suffer from the lisp syndrome that a lot of C code is plauged with.
CHILL also has an EVER keyword for loop control: You can write: DO FOR EVER ... and the semantics are very clear. Embedded code often has everlasting loops. From old CHILL experience, I added a "#define ever ;;" so that I could write "for (ever)". One of the youngsters (to quote Bill Bryson: "I don't want to say that he was awfully young, but he was wearing a Cub Scout uniform") got really mad when he discovered this, went through the entire code repository for the project and edited it back to the proper way of making an infinte loop: "while (1)", with a very nasty remark in the SVN repo about "some people" who are not taking things seriously but add funny jokes in the program code...
Which brings us back to the starting point: Recognition. This youngster immediately recognized "while (1)" as a loop that would run for ever; he did not recognize "for (ever)" as such.
For those of us who grew up with Pascal and CHILL and the like, some of those C coding style rules seem to serve more as "Don't you, common man, believe that this is something you will read as anything resembling a prosa description of what the computer will do! We are trained to recognize it, but we have taken measures so that you will not recognize it!" That is one of our tools to stay in power.
|
|
|
|
|
Your first example has one great advantage: When your productivity is measured in number of code lines produced, it wins by a large margin.
The disadvantage is that to get an overview over even a fairly trivial function that really should fit in a screenful, you have to flip back and forth though a pile of pages. In any sufficiently fancy editor you can split the window so you can correlate various parts of the function logic, but then you can fit half as much logic in each tile, and that may be too little!
A second disadvantage: With all that whitespace (frequently, programmers add at least one blank line before and after every loop, after function initialization, ... everywhere!), you have to look really close to see where this function ends and the next function starts. Except that those blank line lovers usually also add a huge comment block before every function to explain that ThePriceOfApplesPerKilogram argument is the price of apples per kilogram, that pi has the value of 3.1415926535897932384626433 and similar essential information.
To some of these programmers, "readability" is a synonym for "whitespace percentage". Sort of homeopathic programming: The more diluted, the more powerful.
On the other hand: The APL ideal of "There is no programming task so complex that it cannot be solved in a single line of APL" goes too far for my taste. An example from Wikipedia -The game of Life:
life←{↑1 ⍵∨.∧3 4=+/,¯1 0 1∘.⊖¯1 0 1∘.⌽⊂⍵}
I put most regex in the same category. Geek & Poke: Yesterday's regex[^]
I'd like to place myself in the middle of the road, but I cannot deny that those homeopatic programmers raises such protests in me that I lean somewhat over to the other side. If you cannot fit every function in its entirety on a single screenful, then you have used too much whitespace.
|
|
|
|
|
This is the first time I have seen "this" being used to prefix a parameter, and Googling doesn't find it. Could you explain, please?
|
|
|
|
|
|
|
Paragraph level.
Are you thinking that the compiler does something more optimized because you wrote it as a sentence instead of paragraph? Or were you thinking that less typing benefits the project overall? Or perhaps some other underlying reason?
Paragraph style is not only more readable and understandable to another programmer reading it for the first time, it is more readable and understandable to you 6 months, a year, or 2 years later when you revisit the code for the first time since you finished it.
Paragraph style is also easier to debug and see runtime values.
When you are not the only programmer involved now or conceivably in the future, paragraph style would have a positive effect on the maintenance portion of lifecycle cost.
Another benefit of paragraph style is that it lends itself better to adding comments that explain the what and why of what you are doing.
modified 1-Jun-18 10:53am.
|
|
|
|
|
When I started I was actually thinking more about separating concepts and mechanics, or maybe better, business vs technology. But I think I lost that some when playing with creating my examples. )
In business applications (vs scientific/engineering) most methods implement some part of a business function.
At an absurdum extreme you could write the application in one procedural block with thousands of lines of branching/conditional code, where each line does no more than a single thing. It contains an atomic concept. So, you wouldn't write if(x == 1 || x == 3) , that is two concepts. At this extreme, every atomic thought is crystal clear, but the big-picture of what the method does becomes a forest-trees thing.
We decompose software into logical, understandably named chunks, like methods and classes. At another absurd extreme, we could make each concept a method - bool IsValueEqualToOne(int x) {if x ==1 return true else return false} . This is just as bad. We have only translated from a technical representation (x == 1)? to a business representation IsValueEqualToOne .
In real-world programming we balance the size and form of our chunks. In some cases, it is better to combine at a "technical" level - `if(x == 1 || x == 3)`. In others, a business level is better
if(AccountHasFunds(sourceAccount, amount)){
debit(sourceAccount, amount);
credit(targetAccount amount);
}
At the opposite extreme or "can't see the forest for the trees" is "needle in a haystack", one liners. Like the examples I gave. ( and tried to disclaim as how I would actually do it. )
So, to your question Quote: compiler does something more optimized... Or perhaps some other underlying reason?
In rereading it sounds like I may be advocating "one-liners", but not really. All the LINQ and stuff has a greater chance of compiler confusion than optimization. Maybe the best I can think of is that each method we write contains a business purpose and a technical implementation. The clearest code communicates both, and the best code does this in a maintainable way.
With LINQ style and functional becoming popular I was curious to see if people were writing more that way.
My preferred method of coding combines approaches, made easier with local functions. (I just make up "transaction syntax for simplicity)
void TransferFunds(source, target, amount){
if(AccountHasFunds(sourceAccount, amount)){
transaction {
debit(sourceAccount, amount);
credit(targetAccount amount);
}
success: {}
rollback: {}
} else { CancelTransfer(); }
[Description("Validates account balance equals or exceeds amount")]
bool AccountHasFunds(string account, decimal amount){
var accountBalance = account.GetBalance();
if ( accountBalance >= amount){
return true;
else {
return false;
}
}
[Description("Debits amount from account")]
void Credit(…){
…
}
...
}
|
|
|
|
|
In my first university level programming course (I had been fiddeling a little around before that), the professor provided one guideline that I still think is great: A program should fit in a single page. Usually, it will be more than half a page, but never more than that. ("One page" was understood to be 60 code lines.)
You can't solve the program in 60 lines of Pascal, you say? OK, so make up another programming language in which you can solve the problem in 60 lines. Say, if an essential part of your solution is to sort a table of records backwards on the surname field, according to Swedish alphabetic sorting, then pretend that you have a language in which "sorting a table backwards on the surname field according to Swedish sorting rules" is a language primitive. When you have made a clean, readable solution in this hypothetical language, you go ahead to make another program by the same basic rules: One for doing that sorting...
I think this is a very good philosophical approach to code modularization: Make yourself a language that lets you solve the problem in 30-60 lines. Less than 30 lines provides so little detail that it is useless (like "void main(void) {solvetheproblem();}"). Make sure that the way the problem is solved is visible in the code! But don't bring in too many details; requiring more than 60 distinct operations (aka. code lines) indicates that you did not use the right language for the task. So make yourself a better suited language.
Obviously, these are main rules, not absolute. Some problems require less than 30 lines, and in any production setting you cannot expect library functions to suit your problem ideally, adding red tape. But as a way of thinking, it is a great approach: Every level of your code should add a significant, but not an excessive amount of detail. And, relate that stepwise detailing to the problem at hand, not to the libraries you use, language or other coding elements.
Even today I think this lesson, in my very first year of study to become a programmer, is one of the most valuable ones I has throughout my studies.
|
|
|
|
|
In our world, and our code reviews, this one line stuff has to go.
The coding standards should dictate the formatting.
BUT: The Situation can dictate exceptions. We have defensible and in-defensible deviations.
Blocks of similar code where reading one line is effectively the same in the block, we allow
the one line to create a "tabular" view of what is transpiring.
BTW, I don't consider breaking down the code into properly formatted statements to be dumbing down the code. I also think that different companies CAN AND SHOULD have different coding standards!
But overall, I believe if you can pick up a piece of code, and not know if YOU wrote it, or someone else on your team wrote it... Then you are cooking with gas.
|
|
|
|
|
Quote: string Foo(int x){
if (x == 1) {
return "one";
} else {
return "not one";
}
}
string Foo(int x) { return x == 1 ? "one" : "not one"; }
(I also tend to twitch when I see a "return" followed by an "else".)
"(I) am amazed to see myself here rather than there ... now rather than then".
― Blaise Pascal
modified 1-Jun-18 12:10pm.
|
|
|
|
|
I've seen a lot of discussion lately about single point of exit vs multiple points of exit. Since my IDE makes them easy to spot I've started using multiples more in if and switch . But when I see the code in a plain text editor it still can look a little confusing.
|
|
|
|
|
I have no issue with "multiple exits"; when used to avoid lengthy "if ... else".
The point is, the "else" in "if ... return else" is totally redunant.
And sometimes, it's "clearer" to simply return with a simple "if NOT ... return" instead of a lengthy "positive" if.
But "if NOT's" can get confusing if misused.
"(I) am amazed to see myself here rather than there ... now rather than then".
― Blaise Pascal
|
|
|
|
|
Got it. See what you are saying.
|
|
|
|
|
And sometimes I do:
if (... ) {
} else {
return;
}
to avoid a confusing "if NOT" or a lengthy "code block".
"(I) am amazed to see myself here rather than there ... now rather than then".
― Blaise Pascal
|
|
|
|
|
I frequently use multiple exits, to simplify the code structure, reducing nesting levels.
However, the last bug I fixed in one program that is my responsibility was related to this: A few months ago, I added a "try / catch / finally" to improve error handling. Within that "try" was an age old "return" for a rather curious case, hit only if the PC did not have a sufficiently updatet dotNet (and since updates are pushed from a central deployment server, this only happens if you plug in a portable that is not yet detected by the deployment server). Before the "return" was executed, the "finally" block was processed, which was certainly not appropriate in that situation.
I will certainly not stop sprinkling my code with return statements, but I have learned a lesson about paying attention to finally-clauses.
|
|
|
|
|
If God wanted us to program in sentences, He would have abolished all languages except APL.
|
|
|
|
|
Would you call an Italian hooker a pasta-tute?
Bad command or file name. Bad, bad command! Sit! Stay! Staaaay...
AntiTwitter: @DalekDave is now a follower!
|
|
|
|
|