fix WSclient_t initialization in server

[d-a-v]

Reference: discussion with @mcspr

(opened to discussion or any other solution)

[d-a-v]

If the N WSclient_t instances from the array do not need to be re-initialized, then in-place destructor and constructor are not needed. WSclient_t members are always initialized to their default values (0 for scalar, respective contructor for other members including the Strings - this initialization was later destroyed by the former memset()).
Then a WSclient_t::init(num, interval/count...) is sufficient, and would come after the constructors are called (which happens at the time of the array allocation).

[mcspr]

Then a WSclient_t::init(num, interval/count...) is sufficient, and would come after the constructors are called (which happens at the time of the array allocation).

Basically, comes down to this.

Note that classname() = default; does not always guarantee that it is called :) e.g.

Object obj;

vs.

Object obj = Object();

Which is exactly the issue the default values in the class body always help to avoid (and I really like that option tbh). And not to mention different ctors, which need to manually call the default one.

Just to repeat myself - explicit dtor + ctor through placement new could be replaced with either copy-assignment or move-assignment, depending on what we want with the resulting objects.
e.g. like to preserve existing buffers of Strings, which what is currently happening after String="", since it just sets length to 0 and leaves the existing buffer.
Might also be less scary to look at.

[mcspr]

Small code example re. "And not to mention different ctors, which need to manually call the default one."
https://godbolt.org/z/rrY9jP vs. https://godbolt.org/z/GqbdhT (adding default body)

[d-a-v]

edited2

Well OK, thanks!
I just proved myself through another example that scalars are not initialized by default constructor in a placement new
but members with constructors are (so scalars have no implicit constructors).

So basically a new is a calloc() followed by (calling constructors) = (in-place new).
Calling an in-place constructor will not call calloc again.
I was wrongly thinking new was a malloc, then (memset + calling constructors) = (in-place new).

[d-a-v]

@mcspr Do you agree with the last commit ?

(I wouldn't mind if you came with your own PR/solution - the memset() needs to be fixed)

[mcspr]

(ref. gitter + 9d9f892)
I think this covers it.

init() alternative is also doing it like https://godbolt.org/z/zaTjvT
ping intervals are dynamic though, but it seems to be handled via enableHeartbeat

[d-a-v]

@Links2004 This PR is ready for reviewing, sorry for the noise ;)

[Links2004]

no worry, the git commends are for discussing.
and I am happy to learn new stuff and getting the lib better in the process.