It is possible that some of the DFA states constructed using the
previous algorithm will contain items which suggest both ``shift'' and
``reduce'' actions. This situation is known as a shift/reduce
conflict.
Because we have conflicts in the handle-recognising DFA, the parser
will not know which action to take; it will have to pick one
arbitrarily and try it out. Because of this non-determinism, we say
that the grammar was not 
.
An alternative problem could arise if we have two or more items in a
state which have the ``
'' marker to the right of their
r.h.s.; thus we would not know which one to reduce by. This situation
is called a reduce/reduce conflict. Instances of this are rarer
than shift/reduce conflicts.
[Note that no state can have a shift/shift conflict since we are
using a DFA; this sort of problem would arise only as a result of
using an NFA.]
Any state in a handle-recognising DFA which contains either a
shift/reduce or a reduce/reduce conflict is known as an inadequate state.
It is the ideal of 
parser design to minimise the number of
inadequate states without sacrificing efficiency. The most common way
of doing this is to introduce lookahead into the parser construction,
leading to the 
family of parsers.