Re: String Buffer

["Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de>]

On 2021-12-02 22:06, Jeffrey R.Carter wrote:
> On 2021-12-02 21:15, Dmitry A. Kazakov wrote:
>>
>> It is true only to a certain degree. When comparing algorithms it is 
>> valid to claim inefficiency without any context if computational 
>> complexity sufficiently differs.
>>
>> For example, an O(N) algorithm is unquestionably inefficient comparing 
>> to O(log N) one. I leave marginal cases of very small N.
> 
> This is false. If you have the O(N) algorithm and it meets your 
> requirements, then using it is more efficient than implementing the 
> O(log N) algorithm.

You have to show that it meets these requirements. The burden of proof 
is on you as you select an objectively inferior algorithm. Selecting a 
better algorithm is a safe choice under ill-defined conditions.

> Unbounded_String is needed far less than many people think, but there 
> are application domains where it is much easier to achieve correctness 
> and clarity with a variable-length string abstraction than without. 
> Blanket statements about "efficiency" are dangerous for those working in 
> such domains.

It is a good argument because introducing String requires more initial 
efforts in Ada than a thoughtless application Unbounded_String.

And note, that in most cases it is really thoughtless as the choice is 
made on the basis of how easy it is to declare a string component of a 
record type and then rewrite it.

Later on throughout the rest of the program the user of Unbounded_String 
will be consistently punished for that poor choice because normal string 
operations are very uncomfortable with Unbounded_String. But that 
happens later. Right now and here, let us save a couple of code lines.

So the simplest and most persuasive blanket statement is OK to dissuade 
people from poor choices.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de