[boost] Re: BGL: more doc buglets

From: Vladimir Prus (ghost_at_[hidden])
Date: 2003-01-29 02:15:00

Hi Jeremy,

> On Sun, 5 Jan 2003, Vladimir Prus wrote:
> ghost>
> ghost> I've come across more problems with documentation.
> ghost>
> ghost> 1. The docs for topological_sort say that if (u,v) edge is present,
> ghost>     then u comes before v in the topological order. I was assuming
> ghost>     that if I store the order in vector, then u will be found before
> ghost>     v. Instead, the order is reversed! This can only be learned from
> ghost>     the example at the botton -- which is the last place to look at.
> 
> Yes.

Fixed.

> ghost> 2.  The example for the same function is wrong. The graph has a
> ghost>       5 -> 5 edge and is not DAG. libs/graph/example/topo_sort.cpp
> ghost>      fails for that reason.
> 
> Right.

Fixed.

> ghost> 3.  Docs the the same function say that the default value of i_map
> ghost>      parameter is "get(vertex_index, g)". Unless I'm missing something,
> ghost>      this should be "get(vertex_index_t(), g)".
> 
> No, I think vertex_index is fine. vertex_index_t is an enum with one
> value in it, namely vertex_index. See properties.hpp.

You are right. However, docs say only about

    struct vertex_index_t { };

(when discussing adjacency_list), so there's still some inconsistency.

> 
> ghost> 4.  The docs for the "write_graphviz" function do not mention the
> ghost>      "default_writer" class. I believe they should --- it's important
> ghost>      when one want to output onle edge properties.
> 
> Yes.

Edited the docs. Could you please check if I've spoiled anything?

> ghost> 5.   libs/graph/doc/PropertyGraph.html says:
> ghost>
> ghost>      boost::property_map<G, PropertyTag>::type
> ghost>      The type of the property map for the property specified by PropertyTag.
> ghost>      This type must be a model of ReadWritePropertyMap with a key type the same
> ghost>      as the graph's vertex or (****) descriptor type.
> ghost>
> ghost>      "edge" is missing in the marked position.
> 
> Right.

Fixed.

> ghost> 6.   I did not see anywhere the stament that property map obtained from
> ghost>       PropertyGraph are actually a kind of references into the internal
> ghost>       property map. I can conjecure this is so, but a clean statement would
> ghost>       we better.
> 
> Sure.

Do you think it's better to say that all property map have this shared copy 
semantics, or only those from PropertyGraph?

- Volodya