Re: [boost] [Review] ITL review starts today, February 18th

Subject: Re: [boost] [Review] ITL review starts today, February 18th
From: vicente.botet (vicente.botet_at_[hidden])
Date: 2010-02-28 12:40:47

----- Original Message ----- 
From: "Hartmut Kaiser" <hartmut.kaiser_at_[hidden]>
To: <boost_at_[hidden]>; <boost-announce_at_[hidden]>; <boost-users_at_[hidden]>
Sent: Thursday, February 18, 2010 2:22 PM
Subject: [boost] [Review] ITL review starts today, February 18th

> The formal review of the Interval Template Library (ITL) starts today,
> February 18th, 2010 and will end February 27th, 2010. 
> ITL is being developed by Joachim Faulhaber.
> 
> ------------------------------------------------

Hi,  
> - What is your evaluation of the design?

I think the library must made a clear distinction between continuous and discrete intervals, as we are unable to iterate on continuous interval_set using the element iterators.

As other have suggested, the template parameter domain must follow some basic interval requirements, in particular the separation between compile time and run-time bound checking.

I have taken some time to undesrstand the subtilities of the different interval combining styles.  Once understood I think all these are useful alternatives.

> - What is your evaluation of the implementation?

Not reviewed.

> - What is your evaluation of the documentation?

Even if the examples show a lot of things, a tutorial will allow you to explain how to do specific things, starting from basic uses cases and going to more advanced ones.

As noted off-list, the hint parameter of the map insert function is missing.

> - What is your evaluation of the potential usefulness of the library?

In particular I had two uses cases of this library
* Storage of sparse free local unique identifiers (see Boost.LUID on the Sandbox).
* Map memory areas (representing variables) to the transactional cache (in Boost.STM). The memory areas can not overlap as the represent
the memory occuped by variables, but a memory area can be embedded on another, the memory of a field is embedded
on the struct variable.

The variable stored in the maps contains already the size of the variable, and the interval are all of the same type, so the run-time check about the bound types could be avoided in this case. 

Both use cases can be implemented using the library. 

> - Did you try to use the library?  With what compiler?  Did you have any
> problems?

I have tryed to implemented the map memory area without not too much issues, but I had no time to comile and test :(.

> - How much effort did you put into your evaluation? A glance? A quick
> reading? In-depth study?

I've followed the library evolution since its first announcement, and I have do a quick reading of the last documentation.

> - Are you knowledgeable about the problem domain?

I have never need to work with continuous intervals, so I don't knwo the subtilities of this doamin. However I have already 
implemented maps and set of intervals and I find that Boost.ITL do this task much better than my in house implementation.

+1/2. I consider that Boost.ITL must be accepted provisionaly and should be accepted after a fast review once 
the modification I and others have suggest are taken in account.

Best,
Vicente