So, are all your functions named thisFunctionReturnsDouble()? I'd like to see your code :p

If v is double and you change the function to something that returns int, v becomes int. And somewhere else, where you where doing other_var=v/33;, you get something different of what you expect. But no errors.

Why did you have var = v / 33; here, but not var = v / 33.0; if like so much too take precisions? Why don't you write var = double(v) / 33.0;

There is no memory leak and the program works. But not as expected

This? http://blog.johnmuellerbooks.com/2013/04/12/resetting-your-codeblocks-configuration/

Lets return to this example. If you have reasons to assume this function returns integer, I dont understand why it can be changed to double. If not, you should have performed explicit cast yourself

Lets return to this example. If you have reasons to assume this function returns integer, I dont understand why it can be changed to double. If not, you should have performed explicit cast yourself

auto t=myTime(); I'm returning a time. In what format? Integer number of seconds, float number of seconds, a struct? auto t=myTimeInSeconds(); Is it an int or a float, or a double? Or did the programmer created a mySeconds class?

How it's for me: I have warnings for implicit types casts (yes, I cast everything, even int to double with static_cast myself). There are lots of overloaded and templated functions that take different types (e.g. min(T a, T b) or to_string(T). If I'll put auto everywhere, I can get lot of places where it can change some logic if I accidentally change type somewhere

God, who said its better to use auto in this oarticular case?

I have warns for sign mismatch and use braces

How it will help for example when I change a function to return double instead of int and then call some overload function that can take both with the result?

You are asking me for examples. This is one. Obviously I can't give you all and every case where auto leads to problems

Again, I don't say "never use auto", but in most cases, I don't think the risks compensate the usually small advantages

Which function can suddenly return int instead of double?

Or meters, or any kind of measurement

I gave you one example. You have a function that returns a number of seconds (or minutes, or hours). At first it's fine that it returns an int, and then you realize it should be a double

Oh, you asked me, which name I use for a function. Right now, I am working with function shared_ptr<fucking_rabbit_very_fast_random_move_class> next_step()

Or following @Pavel's example. If you have template specialization, suddenly you may end up using a completely different template

Which is mostly bad architecture

Cool. I have no idea what methods fucking_rabbit have, and what do they return

So template specialization is bad architecture...

Which is mostly bad architecture

Emm, function name is next_step(), and return type is shared_ptr<fucking_rabbit_very_fast_random_move_class>

Maybe bad, but it's everywhere. Even using just STL and some popular parsers we can get some problems like that.

I can name lots of good specislizations in std and couple of nasty ones

I dont need to. Most specializations work well with auto. Some dont

Well, so now, several thousands lines away, you have auto v=next_step();

Ok, if you work in an environment where all the code is 100% tested and all the programmers are geniuses, I envy you

If they dont work together Id write a typedef

typedef makes code unreadeble

Yeah, and? Firstly, I make my code more readeble, becouse, I don't need to write 150 symblos, just to initialize some variable, secondly, I write code faster, becouse I don't make misstake in naming, so IDE can autocomplete my code, and so on and son on

typedef makes code unreadeble

You should use using for type aliases, not typedef

They are equivalent

So, lets typedef long long ll; And that, who try to understand what you've wrote will confuse

No, they aren't

No, they aren't

I definitely didnt mean this usage

So in which case typedef is better, than good naming?

using supports template parameters and looks better, but are there other differences?

Isn't it enough differences?

Emm, function name is next_step(), and return type is shared_ptr<fucking_rabbit_very_fast_random_move_class>

Read a book

Read a book

Please which book?

Stephen Prata

with namespaces it is about 67 symbols, so I use auto

I would use ctrl+c, ctrl+v and if you change the return type to unique_ptr and forgot about it, you get a compile error, instead of a runtime error

I would use ctrl+c, ctrl+v and if you change the return type to unique_ptr and forgot about it, you get a compile error, instead of a runtime error

cntr+c, cntr+v from where? So I need to navigate to file, where I created a class, paste it 3 times, make 200 symbols line, just becouse I don't want to use auto. It is stupid

It can make code less readable

You have to navigate the same to find out what auto means. The difference is that with copy-paste, you do it once. With auto, every time you revisit the code