X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fexpressions%2Fparse.c;h=22c7a410802d2fb4556f7cfa0a541b90b34f3e20;hb=faef2451e5aaeb4841a7b8a42a0e8386e591d257;hp=906423650753a45d3b3fd9328b82858de57f6ce2;hpb=96994a54e60e9c95b8bba54c2281acf7059b1203;p=pspp diff --git a/src/language/expressions/parse.c b/src/language/expressions/parse.c index 9064236507..22c7a41080 100644 --- a/src/language/expressions/parse.c +++ b/src/language/expressions/parse.c @@ -59,51 +59,88 @@ atom_type expr_node_returns (const union any_node *); static const char *atom_type_name (atom_type); static struct expression *finish_expression (union any_node *, struct expression *); -static bool type_check (struct expression *, union any_node **, - enum expr_type expected_type); +static bool type_check (const union any_node *, enum val_type expected_type); static union any_node *allocate_unary_variable (struct expression *, const struct variable *); /* Public functions. */ -/* Parses an expression of the given TYPE. - If DICT is nonnull then variables and vectors within it may be - referenced within the expression; otherwise, the expression - must not reference any variables or vectors. - Returns the new expression if successful or a null pointer - otherwise. */ +/* Parses an expression of the given TYPE. If DS is nonnull then variables and + vectors within it may be referenced within the expression; otherwise, the + expression must not reference any variables or vectors. Returns the new + expression if successful or a null pointer otherwise. */ struct expression * -expr_parse (struct lexer *lexer, struct dataset *ds, enum expr_type type) +expr_parse (struct lexer *lexer, struct dataset *ds, enum val_type type) { - union any_node *n; - struct expression *e; + assert (val_type_is_valid (type)); - assert (type == EXPR_NUMBER || type == EXPR_STRING || type == EXPR_BOOLEAN); + struct expression *e = expr_create (ds); + union any_node *n = parse_or (lexer, e); + if (!n || !type_check (n, type)) + { + expr_free (e); + return NULL; + } - e = expr_create (ds); - n = parse_or (lexer, e); - if (n != NULL && type_check (e, &n, type)) - return finish_expression (expr_optimize (n, e), e); - else + return finish_expression (expr_optimize (n, e), e); +} + +/* Parses a boolean expression, otherwise similar to expr_parse(). */ +struct expression * +expr_parse_bool (struct lexer *lexer, struct dataset *ds) +{ + struct expression *e = expr_create (ds); + union any_node *n = parse_or (lexer, e); + if (!n) + { + expr_free (e); + return NULL; + } + + atom_type actual_type = expr_node_returns (n); + if (actual_type == OP_number) + n = expr_allocate_binary (e, OP_NUM_TO_BOOLEAN, n, + expr_allocate_string (e, ss_empty ())); + else if (actual_type != OP_boolean) { + msg (SE, _("Type mismatch: expression has %s type, " + "but a boolean value is required here."), + atom_type_name (actual_type)); expr_free (e); return NULL; } + + return finish_expression (expr_optimize (n, e), e); } -/* Parses and returns an expression of the given TYPE, as - expr_parse(), and sets up so that destroying POOL will free - the expression as well. */ +/* Parses a numeric expression that is intended to be assigned to newly created + variable NEW_VAR_NAME. (This allows for a better error message if the + expression is not numeric.) Otherwise similar to expr_parse(). */ struct expression * -expr_parse_pool (struct lexer *lexer, - struct pool *pool, - struct dataset *ds, - enum expr_type type) +expr_parse_new_variable (struct lexer *lexer, struct dataset *ds, + const char *new_var_name) { - struct expression *e = expr_parse (lexer, ds, type); - if (e != NULL) - pool_add_subpool (pool, e->expr_pool); - return e; + struct expression *e = expr_create (ds); + union any_node *n = parse_or (lexer, e); + if (!n) + { + expr_free (e); + return NULL; + } + + atom_type actual_type = expr_node_returns (n); + if (actual_type != OP_number && actual_type != OP_boolean) + { + msg (SE, _("This command tries to create a new variable %s by assigning a " + "string value to it, but this is not supported. Use " + "the STRING command to create the new variable with the " + "correct width before assigning to it, e.g. STRING %s(A20)."), + new_var_name, new_var_name); + expr_free (e); + return NULL; + } + + return finish_expression (expr_optimize (n, e), e); } /* Free expression E. */ @@ -194,7 +231,7 @@ measure_stack (const union any_node *n, int i; args = *height; - for (i = 0; i < n->composite.arg_cnt; i++) + for (i = 0; i < n->composite.n_args; i++) measure_stack (n->composite.args[i], &args, max); return_height = atom_type_stack (operations[n->type].returns); @@ -247,15 +284,13 @@ finish_expression (union any_node *n, struct expression *e) converted to type EXPECTED_TYPE, inserting a conversion at *N if necessary. Returns true if successful, false on failure. */ static bool -type_check (struct expression *e, - union any_node **n, enum expr_type expected_type) +type_check (const union any_node *n, enum val_type expected_type) { - atom_type actual_type = expr_node_returns (*n); + atom_type actual_type = expr_node_returns (n); switch (expected_type) { - case EXPR_BOOLEAN: - case EXPR_NUMBER: + case VAL_NUMERIC: if (actual_type != OP_number && actual_type != OP_boolean) { msg (SE, _("Type mismatch: expression has %s type, " @@ -263,12 +298,9 @@ type_check (struct expression *e, atom_type_name (actual_type)); return false; } - if (actual_type == OP_number && expected_type == EXPR_BOOLEAN) - *n = expr_allocate_binary (e, OP_NUM_TO_BOOLEAN, *n, - expr_allocate_string (e, ss_empty ())); break; - case EXPR_STRING: + case VAL_STRING: if (actual_type != OP_string) { msg (SE, _("Type mismatch: expression has %s type, " @@ -496,12 +528,12 @@ struct operator On failure, returns false and, if OPERATOR is non-null, sets *OPERATOR to a null pointer. */ static bool -match_operator (struct lexer *lexer, const struct operator ops[], size_t op_cnt, +match_operator (struct lexer *lexer, const struct operator ops[], size_t n_ops, const struct operator **operator) { const struct operator *op; - for (op = ops; op < ops + op_cnt; op++) + for (op = ops; op < ops + n_ops; op++) if (lex_token (lexer) == op->token) { if (op->token != T_NEG_NUM) @@ -516,27 +548,27 @@ match_operator (struct lexer *lexer, const struct operator ops[], size_t op_cnt, } static bool -check_operator (const struct operator *op, int arg_cnt, atom_type arg_type) +check_operator (const struct operator *op, int n_args, atom_type arg_type) { const struct operation *o; size_t i; assert (op != NULL); o = &operations[op->type]; - assert (o->arg_cnt == arg_cnt); + assert (o->n_args == n_args); assert ((o->flags & OPF_ARRAY_OPERAND) == 0); - for (i = 0; i < arg_cnt; i++) + for (i = 0; i < n_args; i++) assert (is_compatible (arg_type, o->args[i])); return true; } static bool -check_binary_operators (const struct operator ops[], size_t op_cnt, +check_binary_operators (const struct operator ops[], size_t n_ops, atom_type arg_type) { size_t i; - for (i = 0; i < op_cnt; i++) + for (i = 0; i < n_ops; i++) check_operator (&ops[i], 2, arg_type); return true; } @@ -555,7 +587,7 @@ get_operand_type (const struct operator *op) one operator/operand pair is parsed. */ static union any_node * parse_binary_operators (struct lexer *lexer, struct expression *e, union any_node *node, - const struct operator ops[], size_t op_cnt, + const struct operator ops[], size_t n_ops, parse_recursively_func *parse_next_level, const char *chain_warning) { @@ -563,11 +595,11 @@ parse_binary_operators (struct lexer *lexer, struct expression *e, union any_nod int op_count; const struct operator *operator; - assert (check_binary_operators (ops, op_cnt, operand_type)); + assert (check_binary_operators (ops, n_ops, operand_type)); if (node == NULL) return node; - for (op_count = 0; match_operator (lexer, ops, op_cnt, &operator); op_count++) + for (op_count = 0; match_operator (lexer, ops, n_ops, &operator); op_count++) { union any_node *rhs; @@ -909,11 +941,22 @@ parse_primary (struct lexer *lexer, struct expression *e) case T_LPAREN: { - union any_node *node; - lex_get (lexer); - node = parse_or (lexer, e); - if (node != NULL && !lex_force_match (lexer, T_RPAREN)) + /* Count number of left parentheses so that we can match them against + an equal number of right parentheses. This defeats trivial attempts + to exhaust the stack with a lot of left parentheses. (More + sophisticated attacks will still succeed.) */ + size_t n = 0; + while (lex_match (lexer, T_LPAREN)) + n++; + + union any_node *node = parse_or (lexer, e); + if (!node) return NULL; + + for (size_t i = 0; i < n; i++) + if (!lex_force_match (lexer, T_RPAREN)) + return NULL; + return node; } @@ -954,7 +997,7 @@ parse_vector_element (struct lexer *lexer, struct expression *e) /* Individual function parsing. */ -const struct operation operations[OP_first + OP_cnt] = { +const struct operation operations[OP_first + n_OP] = { #include "parse.inc" }; @@ -989,61 +1032,53 @@ word_matches (const char **test, const char **name) return true; } +/* Returns 0 if TOKEN and FUNC do not match, + 1 if TOKEN is an acceptable abbreviation for FUNC, + 2 if TOKEN equals FUNC. */ static int -compare_names (const char *test, const char *name, bool abbrev_ok) +compare_function_names (const char *token_, const char *func_) { - if (!abbrev_ok) - return true; - - for (;;) - { - if (!word_matches (&test, &name)) - return true; - if (*name == '\0' && *test == '\0') - return false; - } -} - -static int -compare_strings (const char *test, const char *name, bool abbrev_ok UNUSED) -{ - return c_strcasecmp (test, name); + const char *token = token_; + const char *func = func_; + while (*token || *func) + if (!word_matches (&token, &func)) + return 0; + return !c_strcasecmp (token_, func_) ? 2 : 1; } static bool -lookup_function_helper (const char *name, - int (*compare) (const char *test, const char *name, - bool abbrev_ok), - const struct operation **first, - const struct operation **last) +lookup_function (const char *token, + const struct operation **first, + const struct operation **last) { - const struct operation *f; + *first = *last = NULL; + const struct operation *best = NULL; - for (f = operations + OP_function_first; + for (const struct operation *f = operations + OP_function_first; f <= operations + OP_function_last; f++) - if (!compare (name, f->name, !(f->flags & OPF_NO_ABBREV))) - { - *first = f; + { + int score = compare_function_names (token, f->name); + if (score == 2) + { + best = f; + break; + } + else if (score == 1 && !(f->flags & OPF_NO_ABBREV) && !best) + best = f; + } - while (f <= operations + OP_function_last - && !compare (name, f->name, !(f->flags & OPF_NO_ABBREV))) - f++; - *last = f; + if (!best) + return false; - return true; - } + *first = best; - return false; -} + const struct operation *f = best; + while (f <= operations + OP_function_last + && !c_strcasecmp (f->name, best->name)) + f++; + *last = f; -static bool -lookup_function (const char *name, - const struct operation **first, - const struct operation **last) -{ - *first = *last = NULL; - return (lookup_function_helper (name, compare_strings, first, last) - || lookup_function_helper (name, compare_names, first, last)); + return true; } static int @@ -1061,22 +1096,22 @@ extract_min_valid (const char *s) static atom_type function_arg_type (const struct operation *f, size_t arg_idx) { - assert (arg_idx < f->arg_cnt || (f->flags & OPF_ARRAY_OPERAND)); + assert (arg_idx < f->n_args || (f->flags & OPF_ARRAY_OPERAND)); - return f->args[arg_idx < f->arg_cnt ? arg_idx : f->arg_cnt - 1]; + return f->args[arg_idx < f->n_args ? arg_idx : f->n_args - 1]; } static bool -match_function (union any_node **args, int arg_cnt, const struct operation *f) +match_function (union any_node **args, int n_args, const struct operation *f) { size_t i; - if (arg_cnt < f->arg_cnt - || (arg_cnt > f->arg_cnt && (f->flags & OPF_ARRAY_OPERAND) == 0) - || arg_cnt - (f->arg_cnt - 1) < f->array_min_elems) + if (n_args < f->n_args + || (n_args > f->n_args && (f->flags & OPF_ARRAY_OPERAND) == 0) + || n_args - (f->n_args - 1) < f->array_min_elems) return false; - for (i = 0; i < arg_cnt; i++) + for (i = 0; i < n_args; i++) if (!is_coercible (function_arg_type (f, i), &args[i])) return false; @@ -1085,34 +1120,32 @@ match_function (union any_node **args, int arg_cnt, const struct operation *f) static void coerce_function_args (struct expression *e, const struct operation *f, - union any_node **args, size_t arg_cnt) + union any_node **args, size_t n_args) { int i; - for (i = 0; i < arg_cnt; i++) + for (i = 0; i < n_args; i++) type_coercion_assert (e, function_arg_type (f, i), &args[i]); } static bool -validate_function_args (const struct operation *f, int arg_cnt, int min_valid) +validate_function_args (const struct operation *f, int n_args, int min_valid) { - int array_arg_cnt = arg_cnt - (f->arg_cnt - 1); - if (array_arg_cnt < f->array_min_elems) - { - msg (SE, _("%s must have at least %d arguments in list."), - f->prototype, f->array_min_elems); - return false; - } + /* Count the function arguments that go into the trailing array (if any). We + know that there must be at least the minimum number because + match_function() already checked. */ + int array_n_args = n_args - (f->n_args - 1); + assert (array_n_args >= f->array_min_elems); if ((f->flags & OPF_ARRAY_OPERAND) - && array_arg_cnt % f->array_granularity != 0) + && array_n_args % f->array_granularity != 0) { - if (f->array_granularity == 2) - msg (SE, _("%s must have an even number of arguments in list."), - f->prototype); - else - msg (SE, _("%s must have multiple of %d arguments in list."), - f->prototype, f->array_granularity); + /* RANGE is the only case we have so far. It has paired arguments with + one initial argument, and that's the only special case we deal with + here. */ + assert (f->array_granularity == 2); + assert (n_args % 2 == 0); + msg (SE, _("%s must have an odd number of arguments."), f->prototype); return false; } @@ -1121,26 +1154,19 @@ validate_function_args (const struct operation *f, int arg_cnt, int min_valid) if (f->array_min_elems == 0) { assert ((f->flags & OPF_MIN_VALID) == 0); - msg (SE, _("%s function does not accept a minimum valid " - "argument count."), f->prototype); + msg (SE, _("%s function cannot accept suffix .%d to specify the " + "minimum number of valid arguments."), + f->prototype, min_valid); return false; } else { assert (f->flags & OPF_MIN_VALID); - if (array_arg_cnt < f->array_min_elems) - { - msg (SE, _("%s requires at least %d valid arguments in list."), - f->prototype, f->array_min_elems); - return false; - } - else if (min_valid > array_arg_cnt) + if (min_valid > array_n_args) { - msg (SE, _("With %s, " - "using minimum valid argument count of %d " - "does not make sense when passing only %d " - "arguments in list."), - f->prototype, min_valid, array_arg_cnt); + msg (SE, _("For %s with %d arguments, at most %d (not %d) may be " + "required to be valid."), + f->prototype, n_args, array_n_args, min_valid); return false; } } @@ -1150,26 +1176,26 @@ validate_function_args (const struct operation *f, int arg_cnt, int min_valid) } static void -add_arg (union any_node ***args, int *arg_cnt, int *arg_cap, +add_arg (union any_node ***args, int *n_args, int *allocated_args, union any_node *arg) { - if (*arg_cnt >= *arg_cap) + if (*n_args >= *allocated_args) { - *arg_cap += 8; - *args = xrealloc (*args, sizeof **args * *arg_cap); + *allocated_args += 8; + *args = xrealloc (*args, sizeof **args * *allocated_args); } - (*args)[(*arg_cnt)++] = arg; + (*args)[(*n_args)++] = arg; } static void put_invocation (struct string *s, - const char *func_name, union any_node **args, size_t arg_cnt) + const char *func_name, union any_node **args, size_t n_args) { size_t i; ds_put_format (s, "%s(", func_name); - for (i = 0; i < arg_cnt; i++) + for (i = 0; i < n_args; i++) { if (i > 0) ds_put_cstr (s, ", "); @@ -1180,7 +1206,7 @@ put_invocation (struct string *s, static void no_match (const char *func_name, - union any_node **args, size_t arg_cnt, + union any_node **args, size_t n_args, const struct operation *first, const struct operation *last) { struct string s; @@ -1191,12 +1217,12 @@ no_match (const char *func_name, if (last - first == 1) { ds_put_format (&s, _("Type mismatch invoking %s as "), first->prototype); - put_invocation (&s, func_name, args, arg_cnt); + put_invocation (&s, func_name, args, n_args); } else { ds_put_cstr (&s, _("Function invocation ")); - put_invocation (&s, func_name, args, arg_cnt); + put_invocation (&s, func_name, args, n_args); ds_put_cstr (&s, _(" does not match any known function. Candidates are:")); for (f = first; f < last; f++) @@ -1216,8 +1242,8 @@ parse_function (struct lexer *lexer, struct expression *e) const struct operation *f, *first, *last; union any_node **args = NULL; - int arg_cnt = 0; - int arg_cap = 0; + int n_args = 0; + int allocated_args = 0; struct string func_name; @@ -1240,7 +1266,7 @@ parse_function (struct lexer *lexer, struct expression *e) } args = NULL; - arg_cnt = arg_cap = 0; + n_args = allocated_args = 0; if (lex_token (lexer) != T_RPAREN) for (;;) { @@ -1248,13 +1274,13 @@ parse_function (struct lexer *lexer, struct expression *e) && lex_next_token (lexer, 1) == T_TO) { const struct variable **vars; - size_t var_cnt; + size_t n_vars; size_t i; - if (!parse_variables_const (lexer, dataset_dict (e->ds), &vars, &var_cnt, PV_SINGLE)) + if (!parse_variables_const (lexer, dataset_dict (e->ds), &vars, &n_vars, PV_SINGLE)) goto fail; - for (i = 0; i < var_cnt; i++) - add_arg (&args, &arg_cnt, &arg_cap, + for (i = 0; i < n_vars; i++) + add_arg (&args, &n_args, &allocated_args, allocate_unary_variable (e, vars[i])); free (vars); } @@ -1264,7 +1290,7 @@ parse_function (struct lexer *lexer, struct expression *e) if (arg == NULL) goto fail; - add_arg (&args, &arg_cnt, &arg_cap, arg); + add_arg (&args, &n_args, &allocated_args, arg); } if (lex_match (lexer, T_RPAREN)) break; @@ -1276,16 +1302,16 @@ parse_function (struct lexer *lexer, struct expression *e) } for (f = first; f < last; f++) - if (match_function (args, arg_cnt, f)) + if (match_function (args, n_args, f)) break; if (f >= last) { - no_match (ds_cstr (&func_name), args, arg_cnt, first, last); + no_match (ds_cstr (&func_name), args, n_args, first, last); goto fail; } - coerce_function_args (e, f, args, arg_cnt); - if (!validate_function_args (f, arg_cnt, min_valid)) + coerce_function_args (e, f, args, n_args); + if (!validate_function_args (f, n_args, min_valid)) goto fail; if ((f->flags & OPF_EXTENSION) && settings_get_syntax () == COMPATIBLE) @@ -1303,7 +1329,7 @@ parse_function (struct lexer *lexer, struct expression *e) goto fail; } - n = expr_allocate_composite (e, f - operations, args, arg_cnt); + n = expr_allocate_composite (e, f - operations, args, n_args); n->composite.min_valid = min_valid != -1 ? min_valid : f->array_min_elems; if (n->type == OP_LAG_Vn || n->type == OP_LAG_Vs) @@ -1311,7 +1337,7 @@ parse_function (struct lexer *lexer, struct expression *e) else if (n->type == OP_LAG_Vnn || n->type == OP_LAG_Vsn) { int n_before; - assert (n->composite.arg_cnt == 2); + assert (n->composite.n_args == 2); assert (n->composite.args[1]->type == OP_pos_int); n_before = n->composite.args[1]->integer.i; dataset_need_lag (e->ds, n_before); @@ -1339,7 +1365,7 @@ expr_create (struct dataset *ds) e->eval_pool = pool_create_subpool (e->expr_pool); e->ops = NULL; e->op_types = NULL; - e->op_cnt = e->op_cap = 0; + e->n_ops = e->allocated_ops = 0; return e; } @@ -1401,14 +1427,14 @@ is_valid_node (union any_node *n) struct composite_node *c = &n->composite; assert (is_composite (n->type)); - assert (c->arg_cnt >= op->arg_cnt); - for (i = 0; i < op->arg_cnt; i++) + assert (c->n_args >= op->n_args); + for (i = 0; i < op->n_args; i++) assert (is_compatible (op->args[i], expr_node_returns (c->args[i]))); - if (c->arg_cnt > op->arg_cnt && !is_operator (n->type)) + if (c->n_args > op->n_args && !is_operator (n->type)) { assert (op->flags & OPF_ARRAY_OPERAND); - for (i = 0; i < c->arg_cnt; i++) - assert (is_compatible (op->args[op->arg_cnt - 1], + for (i = 0; i < c->n_args; i++) + assert (is_compatible (op->args[op->n_args - 1], expr_node_returns (c->args[i]))); } } @@ -1418,23 +1444,23 @@ is_valid_node (union any_node *n) union any_node * expr_allocate_composite (struct expression *e, operation_type op, - union any_node **args, size_t arg_cnt) + union any_node **args, size_t n_args) { union any_node *n; size_t i; n = pool_alloc (e->expr_pool, sizeof n->composite); n->type = op; - n->composite.arg_cnt = arg_cnt; + n->composite.n_args = n_args; n->composite.args = pool_alloc (e->expr_pool, - sizeof *n->composite.args * arg_cnt); - for (i = 0; i < arg_cnt; i++) + sizeof *n->composite.args * n_args); + for (i = 0; i < n_args; i++) { if (args[i] == NULL) return NULL; n->composite.args[i] = args[i]; } - memcpy (n->composite.args, args, sizeof *n->composite.args * arg_cnt); + memcpy (n->composite.args, args, sizeof *n->composite.args * n_args); n->composite.min_valid = 0; assert (is_valid_node (n)); return n; @@ -1531,15 +1557,15 @@ allocate_unary_variable (struct expression *e, const struct variable *v) const struct operation * expr_get_function (size_t idx) { - assert (idx < OP_function_cnt); + assert (idx < n_OP_function); return &operations[OP_function_first + idx]; } /* Returns the number of expression functions. */ size_t -expr_get_function_cnt (void) +expr_get_n_functions (void) { - return OP_function_cnt; + return n_OP_function; } /* Returns the name of operation OP. */ @@ -1558,7 +1584,7 @@ expr_operation_get_prototype (const struct operation *op) /* Returns the number of arguments for operation OP. */ int -expr_operation_get_arg_cnt (const struct operation *op) +expr_operation_get_n_args (const struct operation *op) { - return op->arg_cnt; + return op->n_args; }