etnms/etnms.tex
changeset 140 31711ca31685
parent 139 13a42b418eab
child 141 c8af86f7558e
equal deleted inserted replaced
139:13a42b418eab 140:31711ca31685
  1652 $r \gg \retrieve \; r \; (r\downarrow) \; \textit{id} \; s \; v) \; c \; v)$\\
  1652 $r \gg \retrieve \; r \; (r\downarrow) \; \textit{id} \; s \; v) \; c \; v)$\\
  1653 \end{tabular}
  1653 \end{tabular}
  1654 \end{center}
  1654 \end{center}
  1655 Here $\gg$ is almost like an $\textit{NFA}$ in the sense that 
  1655 Here $\gg$ is almost like an $\textit{NFA}$ in the sense that 
  1656 it simulates the lexing process with respect to different strings.
  1656 it simulates the lexing process with respect to different strings.
  1657 \end{itemize}
  1657 
  1658 Our hope is that using $\gg$ we can prove the bits
  1658 Our hope is that using $\gg$ we can prove the bits
  1659 information are not lost when we simplify a regular expression,
  1659 information are not lost when we simplify a regular expression,
  1660 so we need to relate $\gg$ with simplifcation, for example, 
  1660 so we need to relate $\gg$ with simplifcation, for example, 
  1661 one of the lemmas we have proved about $\gg$ is that
  1661 one of the lemmas we have proved about $\gg$ is that
  1662  
  1662 \item
  1663 $\simp$
  1663 \begin{center}
       
  1664 $\simp \; a \gg \textit{bs} \iff  a \gg \textit{bs}$
       
  1665 \end{center}
       
  1666 This could be a step towards our goal, as
       
  1667 it assures that after simplification none of the 
       
  1668 bitsequence that can be created by
       
  1669 the original annotated regular expression
       
  1670 is lost.
       
  1671 If we could prove the following that would be
       
  1672 another step towards our proof, 
       
  1673 
       
  1674 \item
       
  1675 \begin{center}
       
  1676 $(\simp \;a) \backslash s \gg \textit{bs} \iff  a\backslash s \gg \textit{bs}$
       
  1677 \end{center}
       
  1678 
       
  1679 as it says
       
  1680 the simplified regular expression after derivatives will 
       
  1681 still have the full capacity of producing bitsequences
       
  1682 as the unsimplified ones-- pretty much
       
  1683 the intuition we try to establish.
       
  1684 And if we could prove
       
  1685 \item
       
  1686 \begin{center}
       
  1687 $a \backslash s \gg \textit{bs} \iff  a\backslash_\textit{simp} s \gg \textit{bs}$
       
  1688 \end{center}
       
  1689 That would be just a stone's throw away
       
  1690 from $\blexer \; r \; s = \blexers \; r \; s$.
       
  1691 
       
  1692 \end{itemize}
       
  1693 What we do after we work out 
       
  1694 the proof of the above lemma 
       
  1695 is still not clear. It is one of the next steps we need to 
       
  1696 work on.
       
  1697 
  1664 \subsection{the $\textit{ders}_2$ Function}
  1698 \subsection{the $\textit{ders}_2$ Function}
  1665 If we want to prove the result 
  1699 If we want to prove the result 
  1666 \begin{center}
  1700 \begin{center}
  1667 	$ \textit{blexer}\_{simp}(r, \; s) =  \textit{blexer}(r, \; s)$
  1701 	$ \textit{blexer}\_{simp}(r, \; s) =  \textit{blexer}(r, \; s)$
  1668 \end{center}
  1702 \end{center}
  1669 inductively 
  1703 inductively 
  1670 on the structure of the regular expression,
  1704 on the structure of the regular expression,
  1671 then we need to induct on the  case $r_1 \cdot r_2$,
  1705 then we need to induct on the  case $r_1 \cdot r_2$,
  1672 it can be good if we could express $(r_1 \cdot r_2) \backslash s$
  1706 it can be good if we could express $(r_1 \cdot r_2) \backslash s$
  1673 in terms of $r_1 \backslash s$ and $r_2 \backslash s$,
  1707 in terms of $r_1 \backslash s_1$ and $r_2 \backslash s_1$,
  1674 and this naturally induces the $ders2$ function,
  1708 where $s_1$ is a substring of $s$.
       
  1709 For this we introduce the $\textit{ders2}$ function,
  1675 which does a "convolution" on $r_1$ and $r_2$ using the string
  1710 which does a "convolution" on $r_1$ and $r_2$ using the string
  1676 $s$.
  1711 $s$.
  1677 It is based on the observation that the derivative of $r_1 \cdot r_2$
  1712 It is based on the observation that the derivative of $r_1 \cdot r_2$
  1678 with respect to a string $s$ can actually be written in an "explicit form"
  1713 with respect to a string $s$ can actually be written in an "explicit form"
  1679 composed of $r_1$'s derivative of $s$ and $r_2$'s derivative of $s$.
  1714 composed of $r_1$'s derivative of $s$ and $r_2$'s derivative of $s$.