%include {
#include "config.h"

#include <assert.h>

#include "dfilter-int.h"
#include "syntax-tree.h"
#include "sttype-field.h"
#include "sttype-slice.h"
#include "sttype-op.h"
#include "sttype-function.h"
#include "sttype-set.h"
#include "drange.h"

#include "grammar.h"

#ifdef _WIN32
#pragma warning(disable:4671)
#endif

static stnode_t *
new_function(dfwork_t *dfw, stnode_t *node);

#define FAIL(dfw, node, ...) dfilter_fail(dfw, DF_ERROR_GENERIC, stnode_location(node), __VA_ARGS__)

DIAG_OFF_LEMON()
} /* end of %include */

%code {
DIAG_ON_LEMON()
}

/* Parser Information */
%name           Dfilter
%token_prefix   TOKEN_
%extra_argument {dfwork_t *dfw}

/* Terminal and Non-Terminal types and destructors */
%token_type                     {stnode_t*}
%token_destructor               {
    (void)dfw;
    stnode_free($$);
}

%default_type                   {stnode_t*}
%default_destructor             {stnode_free($$);}

%type           range_node_list {GSList*}
%destructor     range_node_list {drange_node_free_list($$);}

%type           layer           {GSList*}
%destructor     layer           {drange_node_free_list($$);}

%type           function_params {GSList*}
%destructor     function_params {st_funcparams_free($$);}

%type           set_list        {GSList*}
%destructor     set_list        {set_nodelist_free($$);}

%type           set_element     {GSList*}
%destructor     set_element     {set_nodelist_free($$);}

/* This is called as soon as a syntax error happens. After that, 
any "error" symbols are shifted, if possible. */
%syntax_error {
    if (!TOKEN) {
        dfilter_fail(dfw, DF_ERROR_GENERIC, NULL, "Unexpected end of filter expression.");
        return;
    }
    FAIL(dfw, TOKEN, "\"%s\" was unexpected in this context.", stnode_token(TOKEN));
}

/* When a parse fails, mark an error. This occurs after
the above syntax_error code and after the parser fails to
use error recovery, shifting an "error" symbol and successfully
shifting 3 more symbols. */
%parse_failure {
    dfw->parse_failure = TRUE;
}

/* ----------------- The grammar -------------- */

/* Associativity */
%left TEST_OR.
%left TEST_AND.
%right TEST_NOT.
%nonassoc TEST_ALL_EQ TEST_ANY_EQ TEST_ALL_NE TEST_ANY_NE TEST_LT TEST_LE TEST_GT TEST_GE
            TEST_CONTAINS TEST_MATCHES.
%left BITWISE_AND.
%left PLUS MINUS.
%left STAR RSLASH PERCENT.
%nonassoc UNARY_PLUS UNARY_MINUS.

/* Top-level targets */
sentence ::= expr(X).           { dfw->st_root = X; }
sentence ::= .                  { dfw->st_root = NULL; }

expr(X) ::= relation(R).        { X = R; }
expr(X) ::= arithmetic_expr(E). { X = E; }

/* Logical tests */
expr(X) ::= expr(Y) TEST_AND(T) expr(Z).
{
    X = T;
    sttype_oper_set2(X, STNODE_OP_AND, Y, Z);
}

expr(X) ::= expr(Y) TEST_OR(T) expr(Z).
{
    X = T;
    sttype_oper_set2(X, STNODE_OP_OR, Y, Z);
}

expr(X) ::= TEST_NOT(T) expr(Y).
{
    X = T;
    sttype_oper_set1(X, STNODE_OP_NOT, Y);
}

/* Any expression inside parens is simply that expression */
expr(X) ::= LPAREN expr(Y) RPAREN.  { X = Y; }

/* Entities, or things that can be compared/tested/checked */
atom(A) ::= STRING(S).      { A = S; }
atom(A) ::= CHARCONST(N).   { A = N; }
atom(A) ::= LITERAL(S).     { A = S; }

layer(R) ::= HASH LBRACKET range_node_list(L) RBRACKET.
{
    R = L;
}

layer(R) ::= HASH INTEGER(N).
{
    char *err_msg = NULL;
    drange_node *range = drange_node_from_str(stnode_token(N), &err_msg);
    if (err_msg != NULL) {
        FAIL(dfw, N, "%s", err_msg);
        g_free(err_msg);
    }
    stnode_free(N);
    R = g_slist_append(NULL, range);
}

layered_field(R) ::= FIELD(F).
{
    R = F;
}

layered_field(R) ::= FIELD(F) layer(L).
{
    R = F;
    sttype_field_set_range(R, L);
    g_slist_free(L);
}

layered_field(R) ::= UNPARSED(U) layer(L).
{
    header_field_info *hfinfo = dfilter_resolve_unparsed(dfw, stnode_token(U));
    if (hfinfo == NULL) {
        FAIL(dfw, U, "%s is not a valid field", stnode_token(U));
    }
    R = stnode_new(STTYPE_FIELD, hfinfo, NULL, stnode_location(U));
    stnode_free(U);
    sttype_field_set_range(R, L);
    g_slist_free(L);
}

field(R) ::= layered_field(F).
{
    R = F;
}

field(R) ::= ATSIGN layered_field(F).
{
    R = F;
    sttype_field_set_raw(R, TRUE);
}

field(R) ::= ATSIGN UNPARSED(U).
{
    const char *token = stnode_token(U);
    const stloc_t *loc = stnode_location(U);
    header_field_info *hfinfo = dfilter_resolve_unparsed(dfw, token);
    if (hfinfo == NULL) {
        FAIL(dfw, U, "%s is not a valid field", stnode_token(U));
    }
    R = stnode_new(STTYPE_FIELD, hfinfo, g_strdup(token), loc);
    sttype_field_set_raw(R, TRUE);
    stnode_free(U);
}

reference(R) ::= DOLLAR LBRACE field(F) RBRACE.
{
    /* convert field to reference */
    R = stnode_new(STTYPE_REFERENCE, sttype_field_hfinfo(F), NULL, stnode_location(F));
    sttype_field_set_drange(R, sttype_field_drange_steal(F));
    sttype_field_set_raw(R, sttype_field_raw(F));
    stnode_free(F);
}

reference(R) ::= DOLLAR LBRACE UNPARSED(U) RBRACE.
{
    header_field_info *hfinfo = dfilter_resolve_unparsed(dfw, stnode_token(U));
    if (hfinfo == NULL) {
        FAIL(dfw, U, "%s is not a valid field", stnode_token(U));
    }
    R = stnode_new(STTYPE_REFERENCE, hfinfo, NULL, stnode_location(U));
    stnode_free(U);
}

entity(E) ::= atom(A).          { E = A; }
entity(E) ::= slice(R).         { E = R; }
entity(E) ::= function(F).      { E = F; }
entity(E) ::= field(F).         { E = F; }
entity(E) ::= reference(R).     { E = R; }
entity(E) ::= UNPARSED(U).
{
    const char *token = stnode_token(U);
    const stloc_t *loc = stnode_location(U);
    header_field_info *hfinfo = dfilter_resolve_unparsed(dfw, token);
    if (hfinfo != NULL) {
        E = stnode_new(STTYPE_FIELD, hfinfo, g_strdup(token), loc);
    }
    else {
        E = stnode_new(STTYPE_LITERAL, g_strdup(token), g_strdup(token), loc);
    }
    stnode_free(U);
}

arithmetic_expr(T) ::= entity(N).
{
    T = N;
}

arithmetic_expr(T) ::= PLUS entity(N).  [UNARY_PLUS]
{
    T = N;
}

arithmetic_expr(T) ::= MINUS(M) entity(N). [UNARY_MINUS]
{
    T = M;
    sttype_oper_set1(T, STNODE_OP_UNARY_MINUS, N);
}

arithmetic_expr(T) ::= arithmetic_expr(F) BITWISE_AND(O) arithmetic_expr(M).
{
    T = O;
    sttype_oper_set2(T, STNODE_OP_BITWISE_AND, F, M);
}

arithmetic_expr(T) ::= arithmetic_expr(F) PLUS(O) arithmetic_expr(M).
{
    T = O;
    sttype_oper_set2(T, STNODE_OP_ADD, F, M);
}

arithmetic_expr(T) ::= arithmetic_expr(F) MINUS(O) arithmetic_expr(M).
{
    T = O;
    sttype_oper_set2(T, STNODE_OP_SUBTRACT, F, M);
}

arithmetic_expr(T) ::= arithmetic_expr(F) STAR(O) arithmetic_expr(M).
{
    T = O;
    sttype_oper_set2(T, STNODE_OP_MULTIPLY, F, M);
}

arithmetic_expr(T) ::= arithmetic_expr(F) RSLASH(O) arithmetic_expr(M).
{
    T = O;
    sttype_oper_set2(T, STNODE_OP_DIVIDE, F, M);
}

arithmetic_expr(T) ::= arithmetic_expr(F) PERCENT(O) arithmetic_expr(M).
{
    T = O;
    sttype_oper_set2(T, STNODE_OP_MODULO, F, M);
}

arithmetic_expr(T) ::= LBRACE arithmetic_expr(F) RBRACE.
{
    T = F;
}

/* Relational tests */
cmp_op(O) ::= TEST_ALL_EQ(L).   { O = L; sttype_oper_set_op(O, STNODE_OP_ALL_EQ); }
cmp_op(O) ::= TEST_ANY_EQ(L).   { O = L; sttype_oper_set_op(O, STNODE_OP_ANY_EQ); }
cmp_op(O) ::= TEST_ALL_NE(L).   { O = L; sttype_oper_set_op(O, STNODE_OP_ALL_NE); }
cmp_op(O) ::= TEST_ANY_NE(L).   { O = L; sttype_oper_set_op(O, STNODE_OP_ANY_NE); }
cmp_op(O) ::= TEST_GT(L).       { O = L; sttype_oper_set_op(O, STNODE_OP_GT); }
cmp_op(O) ::= TEST_GE(L).       { O = L; sttype_oper_set_op(O, STNODE_OP_GE); }
cmp_op(O) ::= TEST_LT(L).       { O = L; sttype_oper_set_op(O, STNODE_OP_LT); }
cmp_op(O) ::= TEST_LE(L).       { O = L; sttype_oper_set_op(O, STNODE_OP_LE); }

comparison_test(T) ::= arithmetic_expr(E) cmp_op(O) arithmetic_expr(F).
{
    T = O;
    sttype_oper_set2_args(O, E, F);
}

/* 'a == b == c' or 'a < b <= c <= d < e' */
comparison_test(T) ::= arithmetic_expr(E) cmp_op(O) comparison_test(R).
{
    stnode_t *L, *F;
    /* for now generate it like E O F  STNODE_OP_AND  F P G, later it could be optimized
       or semantically checked (to make a <= b >= c or a == b != c invalid)?
     */

    F = R;
    do {
        ws_assert(F != NULL && stnode_type_id(F) == STTYPE_TEST);
        sttype_oper_get(F, NULL, &F, NULL);
    } while (stnode_type_id(F) == STTYPE_TEST);

    L = O;
    sttype_oper_set2_args(L, E, stnode_dup(F));

    T = stnode_new(STTYPE_TEST, NULL, NULL, NULL);
    sttype_oper_set2(T, STNODE_OP_AND, L, R);
}

relation_test(T) ::= comparison_test(C).    { T = C; }

relation_test(T) ::= entity(E) TEST_CONTAINS(L) entity(F).
{
    T = L;
    sttype_oper_set2(T, STNODE_OP_CONTAINS, E, F);
}

relation_test(T) ::= entity(E) TEST_MATCHES(L) entity(F).
{
    T = L;
    sttype_oper_set2(T, STNODE_OP_MATCHES, E, F);
}

relation_test(T) ::= entity(E) TEST_IN(O) set(S).
{
    T = O;
    sttype_oper_set2(T, STNODE_OP_IN, E, S);
}

relation_test(T) ::= entity(E) TEST_NOT(P) TEST_IN(O) set(S).
{
    T = P;
    sttype_oper_set2(O, STNODE_OP_IN, E, S);
    sttype_oper_set1(T, STNODE_OP_NOT, O);
}

relation(R) ::= relation_test(T).   { R = T; }

relation(R) ::= ANY relation_test(T).
{
    R = T;
    sttype_test_set_match(R, STNODE_MATCH_ANY);
}

relation(R) ::= ALL relation_test(T).
{
    R = T;
    sttype_test_set_match(R, STNODE_MATCH_ALL);
}

set(S) ::= LBRACE set_list(L) RBRACE.
{
    S = stnode_new(STTYPE_SET, L, NULL, NULL);
}

set_list(L) ::= set_element(N).
{
    L = g_slist_concat(NULL, N);
}

set_list(L) ::= set_list(P) COMMA set_element(N).
{
    L = g_slist_concat(P, N);
}

set_entity(N) ::= entity(E).
{
    N = E;
}

set_entity(N) ::= MINUS(M) entity(E).
{
    N = M;
    sttype_oper_set1(N, STNODE_OP_UNARY_MINUS, E);
}

set_entity(N) ::= PLUS entity(E).
{
    N = E;
}

set_element(N) ::= set_entity(X).
{
    N = g_slist_append(NULL, X);
    N = g_slist_append(N, NULL);
}

set_element(N) ::= set_entity(X) DOTDOT set_entity(Y).
{
    N = g_slist_append(NULL, X);
    N = g_slist_append(N, Y);
}

/* Slices */

slice(R) ::= entity(E) LBRACKET range_node_list(L) RBRACKET.
{
    R = stnode_new(STTYPE_SLICE, NULL, NULL, NULL);
    sttype_slice_set(R, E, L);

    /* Delete the list, but not the drange_nodes that
     * the list contains. */
    g_slist_free(L);
}

range_node_list(L) ::= RANGE_NODE(N).
{
    char *err_msg = NULL;
    drange_node *rn = drange_node_from_str(stnode_token(N), &err_msg);
    if (err_msg != NULL) {
        FAIL(dfw, N, "%s", err_msg);
        g_free(err_msg);
    }
    L = g_slist_append(NULL, rn);
    stnode_free(N);
}

range_node_list(L) ::= range_node_list(P) COMMA RANGE_NODE(N).
{
    char *err_msg = NULL;
    drange_node *rn = drange_node_from_str(stnode_token(N), &err_msg);
    if (err_msg != NULL) {
        FAIL(dfw, N, "%s", err_msg);
        g_free(err_msg);
    }
    L = g_slist_append(P, rn);
    stnode_free(N);
}

/* Functions */

%code {
    static stnode_t *
    new_function(dfwork_t *dfw, stnode_t *node)
    {
        const char *name = stnode_token(node);

        df_func_def_t *def = df_func_lookup(name);
        if (!def) {
            FAIL(dfw, node, "Function '%s' does not exist", name);
        }
        stnode_replace(node, STTYPE_FUNCTION, def);
        return node;
    }
}

/* A function can have one or more parameters */
function(F) ::= UNPARSED(U) LPAREN function_params(P) RPAREN.
{
    F = new_function(dfw, U);
    sttype_function_set_params(F, P);
}

/* A function can have zero parameters. */
function(F) ::= UNPARSED(U) LPAREN RPAREN.
{
    F = new_function(dfw, U);
}

function_params(P) ::= arithmetic_expr(E).
{
    P = g_slist_append(NULL, E);
}

function_params(P) ::= function_params(L) COMMA arithmetic_expr(E).
{
    P = g_slist_append(L, E);
}