A named tuple is an object which contains a fixed amount of
named objects of possibly different type. A named tuple comes
with an embedded hash()-function, which requires that the objects
be hashable, and with equality comparison functions. The hash-function
can be automatically called by the Embedded_Hash hashing object.
The named tuples are generated by the macros PASTEL_NAMED_TUPLEx(),
where the x is the number of contained objects. The macros take as
input the name of the tuple type, and the types and names of
the contained objects.
While tuples are implemented generically in C++11, it is still
perhaps clearer to refer to the contained objects by familiar
names and syntax, rather than by std::get<Name>(tuple).
The need for named tuples comes up in the implementation of
data structures, where composite data needs to be used
as a key of an associative data structure (such as std::map).
For example, in LL-parsing, the applied rule is looked up
based on a the current non-terminal symbol, and the current
lookahead symbol. In this case a named tuple would be used
as a key as follows:
PASTEL_NAMED_TUPLE2(Entry,
Symbol, predicted,
Symbol, lookahead);
std::unordered_map<Entry, Rule, Embedded_Hash> llParseTable;
auto iter = llParseTable.find(Entry(predicted, lookahead));
const Entry& entry = iter->first;
entry.predicted;
entry.lookahead;
This is much clearer than using std::pair (e.g. first and
second), or std::tuple (e.g. std::get<Predicted>(entry)).