theory Lexer

  imports Main 

begin

section {* Sequential Composition of Languages *}

definition

  Sequ :: "string set \<Rightarrow> string set \<Rightarrow> string set" ("_ ;; _" [100,100] 100)

where 

  "A ;; B = {s1 @ s2 | s1 s2. s1 \<in> A \<and> s2 \<in> B}"

text {* Two Simple Properties about Sequential Composition *}

lemma Sequ_empty_string [simp]:

  shows "A ;; {[]} = A"

  and   "{[]} ;; A = A"

by (simp_all add: Sequ_def)

lemma Sequ_empty [simp]:

  shows "A ;; {} = {}"

  and   "{} ;; A = {}"

by (simp_all add: Sequ_def)

section {* Semantic Derivative (Left Quotient) of Languages *}

definition

  Der :: "char \<Rightarrow> string set \<Rightarrow> string set"

where

  "Der c A \<equiv> {s. c # s \<in> A}"

definition

  Ders :: "string \<Rightarrow> string set \<Rightarrow> string set"

where

  "Ders s A \<equiv> {s'. s @ s' \<in> A}"

lemma Der_null [simp]:

  shows "Der c {} = {}"

unfolding Der_def

by auto

lemma Der_empty [simp]:

  shows "Der c {[]} = {}"

unfolding Der_def

by auto

lemma Der_char [simp]:

  shows "Der c {[d]} = (if c = d then {[]} else {})"

unfolding Der_def

by auto

lemma Der_union [simp]:

  shows "Der c (A \<union> B) = Der c A \<union> Der c B"

unfolding Der_def

by auto

lemma Der_Sequ [simp]:

  shows "Der c (A ;; B) = (Der c A) ;; B \<union> (if [] \<in> A then Der c B else {})"

unfolding Der_def Sequ_def

by (auto simp add: Cons_eq_append_conv)

section {* Kleene Star for Languages *}

inductive_set

  Star :: "string set \<Rightarrow> string set" ("_\<star>" [101] 102)

  for A :: "string set"

where

  start[intro]: "[] \<in> A\<star>"

| step[intro]:  "\<lbrakk>s1 \<in> A; s2 \<in> A\<star>\<rbrakk> \<Longrightarrow> s1 @ s2 \<in> A\<star>"

(* Arden's lemma *)

lemma Star_cases:

  shows "A\<star> = {[]} \<union> A ;; A\<star>"

unfolding Sequ_def

by (auto) (metis Star.simps)

lemma Star_decomp: 

by (induct x\<equiv>"c # x" rule: Star.induct) 

   (auto simp add: append_eq_Cons_conv)

lemma Star_Der_Sequ: 

  shows "Der c (A\<star>) \<subseteq> (Der c A) ;; A\<star>"

unfolding Der_def

lemma Der_star [simp]:

  shows "Der c (A\<star>) = (Der c A) ;; A\<star>"

proof -    

  have "Der c (A\<star>) = Der c ({[]} \<union> A ;; A\<star>)"  

    by (simp only: Star_cases[symmetric])

  also have "... = Der c (A ;; A\<star>)"

    by (simp only: Der_union Der_empty) (simp)

  also have "... = (Der c A) ;; A\<star> \<union> (if [] \<in> A then Der c (A\<star>) else {})"

    by simp

  also have "... =  (Der c A) ;; A\<star>"

    using Star_Der_Sequ by auto

  finally show "Der c (A\<star>) = (Der c A) ;; A\<star>" .

qed

section {* Regular Expressions *}

datatype rexp =

  ZERO

| ONE

| CHAR char

| SEQ rexp rexp

| ALT rexp rexp

| STAR rexp

section {* Semantics of Regular Expressions *}

fun

  L :: "rexp \<Rightarrow> string set"

where

  "L (ZERO) = {}"

| "L (ONE) = {[]}"

| "L (CHAR c) = {[c]}"

| "L (SEQ r1 r2) = (L r1) ;; (L r2)"

| "L (ALT r1 r2) = (L r1) \<union> (L r2)"

| "L (STAR r) = (L r)\<star>"

section {* Nullable, Derivatives *}

fun

 nullable :: "rexp \<Rightarrow> bool"

where

  "nullable (ZERO) = False"

| "nullable (ONE) = True"

| "nullable (CHAR c) = False"

| "nullable (ALT r1 r2) = (nullable r1 \<or> nullable r2)"

| "nullable (SEQ r1 r2) = (nullable r1 \<and> nullable r2)"

| "nullable (STAR r) = True"

fun

 der :: "char \<Rightarrow> rexp \<Rightarrow> rexp"

where

  "der c (ZERO) = ZERO"

| "der c (ONE) = ZERO"

| "der c (CHAR d) = (if c = d then ONE else ZERO)"

| "der c (ALT r1 r2) = ALT (der c r1) (der c r2)"

| "der c (SEQ r1 r2) = 

     (if nullable r1

      then ALT (SEQ (der c r1) r2) (der c r2)

      else SEQ (der c r1) r2)"

| "der c (STAR r) = SEQ (der c r) (STAR r)"

fun 

 ders :: "string \<Rightarrow> rexp \<Rightarrow> rexp"

where

  "ders [] r = r"

| "ders (c # s) r = ders s (der c r)"

lemma nullable_correctness:

  shows "nullable r  \<longleftrightarrow> [] \<in> (L r)"

by (induct r) (auto simp add: Sequ_def) 

lemma der_correctness:

  shows "L (der c r) = Der c (L r)"

by (induct r) (simp_all add: nullable_correctness)

lemma ders_correctness:

  shows "L (ders s r) = Ders s (L r)"

apply(induct s arbitrary: r)

apply(simp_all add: Ders_def der_correctness Der_def)

done

section {* Lemmas about ders *}

lemma ders_ZERO:

  shows "ders s (ZERO) = ZERO"

apply(induct s)

apply(simp_all)

done

lemma ders_ONE:

  shows "ders s (ONE) = (if s = [] then ONE else ZERO)"

apply(induct s)

apply(simp_all add: ders_ZERO)

done

lemma ders_CHAR:

  shows "ders s (CHAR c) = 

           (if s = [c] then ONE else 

           (if s = [] then (CHAR c) else ZERO))"

apply(induct s)

apply(simp_all add: ders_ZERO ders_ONE)

done

lemma  ders_ALT:

  shows "ders s (ALT r1 r2) = ALT (ders s r1) (ders s r2)"

apply(induct s arbitrary: r1 r2)

apply(simp_all)

done

section {* Values *}

datatype val = 

  Void

| Char char

| Seq val val

| Right val

| Left val

| Stars "val list"

section {* The string behind a value *}

fun 

  flat :: "val \<Rightarrow> string"

where

  "flat (Void) = []"

| "flat (Char c) = [c]"

| "flat (Left v) = flat v"

| "flat (Right v) = flat v"

| "flat (Seq v1 v2) = (flat v1) @ (flat v2)"

| "flat (Stars []) = []"

| "flat (Stars (v#vs)) = (flat v) @ (flat (Stars vs))" 

lemma flat_Stars [simp]:

 "flat (Stars vs) = concat (map flat vs)"

by (induct vs) (auto)

section {* Relation between values and regular expressions *}

inductive 

  Prf :: "val \<Rightarrow> rexp \<Rightarrow> bool" ("\<turnstile> _ : _" [100, 100] 100)

where

 "\<lbrakk>\<turnstile> v1 : r1; \<turnstile> v2 : r2\<rbrakk> \<Longrightarrow> \<turnstile> Seq v1 v2 : SEQ r1 r2"

| "\<turnstile> v1 : r1 \<Longrightarrow> \<turnstile> Left v1 : ALT r1 r2"

| "\<turnstile> v2 : r2 \<Longrightarrow> \<turnstile> Right v2 : ALT r1 r2"

| "\<turnstile> Void : ONE"

| "\<turnstile> Char c : CHAR c"

inductive_cases Prf_elims:

  "\<turnstile> v : ZERO"

  "\<turnstile> v : SEQ r1 r2"

  "\<turnstile> v : ALT r1 r2"

  "\<turnstile> v : ONE"

  "\<turnstile> v : CHAR c"

lemma Prf_flat_L:

  assumes "\<turnstile> v : r" shows "flat v \<in> L r"

using assms

lemma Prf_Stars_append:

  assumes "\<turnstile> Stars vs1 : STAR r" "\<turnstile> Stars vs2 : STAR r"

  shows "\<turnstile> Stars (vs1 @ vs2) : STAR r"

using assms

lemma Star_string:

  assumes "s \<in> A\<star>"

  shows "\<exists>ss. concat ss = s \<and> (\<forall>s \<in> set ss. s \<in> A)"

using assms

apply(induct rule: Star.induct)

apply(auto)

apply(rule_tac x="[]" in exI)

apply(simp)

apply(rule_tac x="s1#ss" in exI)

apply(simp)

done

lemma Star_val:

  assumes "\<forall>s\<in>set ss. \<exists>v. s = flat v \<and> \<turnstile> v : r"

  shows "\<exists>vs. concat (map flat vs) = concat ss \<and> (\<forall>v\<in>set vs. \<turnstile> v : r)"

using assms

apply(induct ss)

apply(auto)

apply (metis empty_iff list.set(1))

by (metis concat.simps(2) list.simps(9) set_ConsD)

lemma L_flat_Prf1:

  assumes "\<turnstile> v : r" 

  shows "flat v \<in> L r"

using assms

lemma L_flat_Prf2:

  assumes "s \<in> L r" 

  shows "\<exists>v. \<turnstile> v : r \<and> flat v = s"

using assms

lemma L_flat_Prf:

  "L(r) = {flat v | v. \<turnstile> v : r}"

using L_flat_Prf1 L_flat_Prf2 by blast

lemma Star_values_exists:

  assumes "s \<in> (L r)\<star>"

  shows "\<exists>vs. concat (map flat vs) = s \<and> \<turnstile> Stars vs : STAR r"

using assms

apply(drule_tac Star_string)

apply(auto)

section {* Sulzmann and Lu functions *}

fun 

  mkeps :: "rexp \<Rightarrow> val"

where

  "mkeps(ONE) = Void"

| "mkeps(SEQ r1 r2) = Seq (mkeps r1) (mkeps r2)"

| "mkeps(ALT r1 r2) = (if nullable(r1) then Left (mkeps r1) else Right (mkeps r2))"

| "mkeps(STAR r) = Stars []"

fun injval :: "rexp \<Rightarrow> char \<Rightarrow> val \<Rightarrow> val"

where

  "injval (CHAR d) c Void = Char d"

| "injval (ALT r1 r2) c (Left v1) = Left(injval r1 c v1)"

| "injval (ALT r1 r2) c (Right v2) = Right(injval r2 c v2)"

| "injval (SEQ r1 r2) c (Seq v1 v2) = Seq (injval r1 c v1) v2"

| "injval (SEQ r1 r2) c (Left (Seq v1 v2)) = Seq (injval r1 c v1) v2"

| "injval (SEQ r1 r2) c (Right v2) = Seq (mkeps r1) (injval r2 c v2)"

| "injval (STAR r) c (Seq v (Stars vs)) = Stars ((injval r c v) # vs)" 

section {* Mkeps, injval *}

lemma mkeps_nullable:

  assumes "nullable(r)" 

  shows "\<turnstile> mkeps r : r"

using assms

by (induct rule: nullable.induct) 

   (auto intro: Prf.intros)

lemma mkeps_flat:

  assumes "nullable(r)" 

  shows "flat (mkeps r) = []"

using assms

by (induct rule: nullable.induct) (auto)

lemma Prf_injval:

  assumes "\<turnstile> v : der c r" 

  shows "\<turnstile> (injval r c v) : r"

using assms

apply(induct r arbitrary: c v rule: rexp.induct)

apply(auto intro!: Prf.intros mkeps_nullable elim!: Prf_elims split: if_splits)

lemma Prf_injval_flat:

  assumes "\<turnstile> v : der c r" 

  shows "flat (injval r c v) = c # (flat v)"

using assms

apply(induct arbitrary: v rule: der.induct)

done

section {* Our Alternative Posix definition *}

inductive 

  Posix :: "string \<Rightarrow> rexp \<Rightarrow> val \<Rightarrow> bool" ("_ \<in> _ \<rightarrow> _" [100, 100, 100] 100)

where

  Posix_ONE: "[] \<in> ONE \<rightarrow> Void"

| Posix_CHAR: "[c] \<in> (CHAR c) \<rightarrow> (Char c)"

| Posix_ALT1: "s \<in> r1 \<rightarrow> v \<Longrightarrow> s \<in> (ALT r1 r2) \<rightarrow> (Left v)"

| Posix_ALT2: "\<lbrakk>s \<in> r2 \<rightarrow> v; s \<notin> L(r1)\<rbrakk> \<Longrightarrow> s \<in> (ALT r1 r2) \<rightarrow> (Right v)"

| Posix_SEQ: "\<lbrakk>s1 \<in> r1 \<rightarrow> v1; s2 \<in> r2 \<rightarrow> v2;

    \<not>(\<exists>s\<^sub>3 s\<^sub>4. s\<^sub>3 \<noteq> [] \<and> s\<^sub>3 @ s\<^sub>4 = s2 \<and> (s1 @ s\<^sub>3) \<in> L r1 \<and> s\<^sub>4 \<in> L r2)\<rbrakk> \<Longrightarrow> 

    (s1 @ s2) \<in> (SEQ r1 r2) \<rightarrow> (Seq v1 v2)"

| Posix_STAR1: "\<lbrakk>s1 \<in> r \<rightarrow> v; s2 \<in> STAR r \<rightarrow> Stars vs; flat v \<noteq> [];

    \<not>(\<exists>s\<^sub>3 s\<^sub>4. s\<^sub>3 \<noteq> [] \<and> s\<^sub>3 @ s\<^sub>4 = s2 \<and> (s1 @ s\<^sub>3) \<in> L r \<and> s\<^sub>4 \<in> L (STAR r))\<rbrakk>

    \<Longrightarrow> (s1 @ s2) \<in> STAR r \<rightarrow> Stars (v # vs)"

| Posix_STAR2: "[] \<in> STAR r \<rightarrow> Stars []"

inductive_cases Posix_elims:

  "s \<in> ZERO \<rightarrow> v"

  "s \<in> ONE \<rightarrow> v"

  "s \<in> CHAR c \<rightarrow> v"

  "s \<in> ALT r1 r2 \<rightarrow> v"

  "s \<in> SEQ r1 r2 \<rightarrow> v"

  "s \<in> STAR r \<rightarrow> v"

lemma Posix1:

  assumes "s \<in> r \<rightarrow> v"

  shows "s \<in> L r" "flat v = s"

using assms

by (induct s r v rule: Posix.induct)

   (auto simp add: Sequ_def)

lemma Posix1a:

  assumes "s \<in> r \<rightarrow> v"

  shows "\<turnstile> v : r"

using assms

lemma Posix_mkeps:

  assumes "nullable r"

  shows "[] \<in> r \<rightarrow> mkeps r"

using assms

apply(induct r rule: nullable.induct)

apply(auto intro: Posix.intros simp add: nullable_correctness Sequ_def)

apply(subst append.simps(1)[symmetric])

apply(rule Posix.intros)

apply(auto)

done

lemma Posix_determ:

  assumes "s \<in> r \<rightarrow> v1" "s \<in> r \<rightarrow> v2"

  shows "v1 = v2"

using assms

proof (induct s r v1 arbitrary: v2 rule: Posix.induct)

  case (Posix_ONE v2)

  have "[] \<in> ONE \<rightarrow> v2" by fact

  then show "Void = v2" by cases auto

next 

  case (Posix_CHAR c v2)

  have "[c] \<in> CHAR c \<rightarrow> v2" by fact

  then show "Char c = v2" by cases auto

next 

  case (Posix_ALT1 s r1 v r2 v2)

  have "s \<in> ALT r1 r2 \<rightarrow> v2" by fact

  moreover

  have "s \<in> r1 \<rightarrow> v" by fact

  then have "s \<in> L r1" by (simp add: Posix1)

  ultimately obtain v' where eq: "v2 = Left v'" "s \<in> r1 \<rightarrow> v'" by cases auto 

  moreover

  have IH: "\<And>v2. s \<in> r1 \<rightarrow> v2 \<Longrightarrow> v = v2" by fact

  ultimately have "v = v'" by simp

  then show "Left v = v2" using eq by simp

next 

  case (Posix_ALT2 s r2 v r1 v2)

  have "s \<in> ALT r1 r2 \<rightarrow> v2" by fact

  moreover

  have "s \<notin> L r1" by fact

  ultimately obtain v' where eq: "v2 = Right v'" "s \<in> r2 \<rightarrow> v'" 

    by cases (auto simp add: Posix1) 

  moreover

  have IH: "\<And>v2. s \<in> r2 \<rightarrow> v2 \<Longrightarrow> v = v2" by fact

  ultimately have "v = v'" by simp

  then show "Right v = v2" using eq by simp

next

  case (Posix_SEQ s1 r1 v1 s2 r2 v2 v')

  have "(s1 @ s2) \<in> SEQ r1 r2 \<rightarrow> v'" 

       "s1 \<in> r1 \<rightarrow> v1" "s2 \<in> r2 \<rightarrow> v2"

       "\<not> (\<exists>s\<^sub>3 s\<^sub>4. s\<^sub>3 \<noteq> [] \<and> s\<^sub>3 @ s\<^sub>4 = s2 \<and> s1 @ s\<^sub>3 \<in> L r1 \<and> s\<^sub>4 \<in> L r2)" by fact+