[lmi] Searching a symbol in code [Was: master c88b192 5/8: Rename member variables]

[Greg Chicares]

On 2018-04-28 16:53, Vadim Zeitlin wrote:
> On Fri, 27 Apr 2018 10:52:49 -0400 (EDT) Greg Chicares <address@hidden> wrote:
[...]
> GC> commit c88b192094b96f66e0ba97af2f22cf8fdab45d05
[...]
> GC> -        :header_(header)
> GC> -        ,width_(width)
> GC> +        :col_header_(header)
> GC> +        ,col_width_(width)
> 
>  Is it really such a good idea to use column_info::col_width_ as the full
> field name?

No, certainly not in the long term--it's nasty.

> I think column_info::width_ was perfectly unambiguous already

'width_' (variable) was already unambiguous, yes--but...
'width'  (accessor) would have been ambiguous in the next commit (f6c8fa6c).

For now, I'm still trying to achieve a full understanding of this code,
and searching for '\<width\>' finds too many commentary lines.

Maybe after I've understood it we can make these names less nasty.

Here are a few random thoughts that occur to me:

 - I've seen others use a prefix 'the_' in situations like this, but I
   think the_width() is even worse than col_width().

 - We could of course use 'get_' and 'set_' prefixes. That's probably
   better than a fragment of the name of the owning class ('col_').

 - However, I try to avoid 'get_' and 'set_' because they're generally
   too verbose--ideally, no prefix would be needed:
     int length()     |  length(int)     | int length_   // ideal
     int get_length() |  set_length(int) | int length_   // not ideal

> and while I understand the argument about not using common words as
> identifiers to enhance grepability, I also think it shouldn't be overblown
> and lead us to conclude that we should use unique but cryptic or unique but
> overlong names everywhere: for better or worse, grep can't be used to
> explore C++ code structure in any case, so making it easier to use it
> shouldn't be the main goal.

Then let me describe exactly how I'm exploring this code, and maybe you
can show me a better way. In vim, I just move the cursor to 'column' and
press '*', then use 'n' and 'N'. I don't really want to type a complex
regex to indicate 'except after //'. What more powerful method am I
missing?

>  But if we do it like this, then I think either the variables should use
> column_ prefix or the class renamed to col_info because using both prefixes
> in immediate vicinity of each other is just a recipe for cognitive
> dissonance.

Yes, I got stuck on this today:

    for(std::size_t col = 0; col < column; ++col)
        {
        x += all_columns().at(col).col_width();
        }

and it just made me think of "колоколоколокола".