Lambda Library - Review

From: Martin Weiser (weiser_at_[hidden])
Date: 2002-03-18 06:00:07

Here are some review comments on the Lambda Library.

General evaluation:
* The BLL is a useful library improving the usability of the STL a lot. 
Given the language restrictions, I think it's a nearly optimal substitute 
for real lambda expressions.

* The library comes with an extensive and well written documentation.

* It is fairly easy to use from the beginning on, even with only a *very* 
quick glance at the documentation. However, there are a couple of 
pitfalls that become clear only when reading the docs (and not from gcc's 
error messages - well, hardly a surprise).

Conclusion:
The Lambda Library should be accepted as a Boost library.

Minor comments:
* For passing lambda functors around in own templated code it can be 
necessary to obtain the return value of a lambda functor for a given set 
of argument types. The necessary type expression can be derived from the 
information provided in 5.3.3 "Function objects as targets / The sig 
template" and dwelling through some compiler error messages, but should 
definitely be documented explicitly. I didn't find such a passage in the 
docs.
Addidionally, a helper traits class for easier access of the return type 
may be a good idea.

* Writing the identity as 0, _1 instead of _1 comes unexpected. Thus 
either placeholders should be full fledged lambda functors, or a design 
rationale why this is not the case should be given in section 5.1 of the 
documentation.

* I was somewhat irritated about the difference between ref/cref and var 
for delayed access to external variables. This should be explained in 
more detail.

* In section 3.1 in class plus_1, the call operator is incorrect (missing 
pair of parentheses, _i instead of i).

Reviewers approach:
I gave the documentationa an almost complete reading, and played around 
with some trivial test cases. Thus I can't comment on the more arcane 
features of the library (exceptions, nested algorithms,...).

No problems (after reading the docs) with gcc 2.95.2.

The only look I took at the implementation was for finding the lambda 
functors' return type (but failed - given the complexity of the task the 
library addresses, this is hardly surprising). Thus I can't comment on 
the implementation's quality.

Finally, having gathered some experience with template metaprogramming 
when writing part of the View Template Library, I must say I'm quite 
impressed by the work of Jaakko and Gary.

Martin

-- 
Dr. Martin Weiser            Zuse Institute Berlin
weiser_at_[hidden]                Scientific Computing
http://www.zib.de/weiser     Numerical Analysis and Modelling