X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fexpressions%2Fparse.c;h=0493ed878bfd98283451eed17b06a66ea88b3409;hb=bc0434f1ced83fc3a5f5ace44cbad3acdc5e3e29;hp=f9461f16bc222fa5f905321c3a22f3b21f27aebc;hpb=173d1687aea88e0e5e1b1d8615ed68ebefb15d08;p=pspp diff --git a/src/language/expressions/parse.c b/src/language/expressions/parse.c index f9461f16bc..0493ed878b 100644 --- a/src/language/expressions/parse.c +++ b/src/language/expressions/parse.c @@ -1,5 +1,5 @@ /* PSPP - a program for statistical analysis. - Copyright (C) 1997-9, 2000, 2006, 2010 Free Software Foundation, Inc. + Copyright (C) 1997-9, 2000, 2006, 2010, 2011, 2012, 2014 Free Software Foundation, Inc. This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by @@ -23,22 +23,24 @@ #include #include -#include "helpers.h" -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include "xalloc.h" +#include "data/case.h" +#include "data/dictionary.h" +#include "data/settings.h" +#include "data/variable.h" +#include "language/expressions/helpers.h" +#include "language/lexer/format-parser.h" +#include "language/lexer/lexer.h" +#include "language/lexer/variable-parser.h" +#include "libpspp/array.h" +#include "libpspp/assertion.h" +#include "libpspp/i18n.h" +#include "libpspp/message.h" +#include "libpspp/misc.h" +#include "libpspp/pool.h" +#include "libpspp/str.h" + +#include "gl/c-strcase.h" +#include "gl/xalloc.h" /* Declarations. */ @@ -57,49 +59,98 @@ 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. If POOL is nonnull, + then destroying POOL will free the expression; otherwise, the caller must + eventually free it with expr_free(). */ struct expression * -expr_parse (struct lexer *lexer, struct dataset *ds, enum expr_type type) +expr_parse (struct lexer *lexer, struct pool *pool, 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 + e = finish_expression (expr_optimize (n, e), e); + if (pool) + pool_add_subpool (pool, e->expr_pool); + return e; +} + +/* Parses a boolean expression, otherwise similar to expr_parse(). */ +struct expression * +expr_parse_bool (struct lexer *lexer, struct pool *pool, 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; + } + + e = finish_expression (expr_optimize (n, e), e); + if (pool) + pool_add_subpool (pool, e->expr_pool); + return 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 pool *pool, struct dataset *ds, + const char *new_var_name) { - struct expression *e = expr_parse (lexer, ds, type); - if (e != NULL) + 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; + } + + e = finish_expression (expr_optimize (n, e), e); + if (pool) pool_add_subpool (pool, e->expr_pool); return e; } @@ -245,15 +296,13 @@ 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, " @@ -261,11 +310,9 @@ type_check (struct expression *e, atom_type_name (actual_type)); return false; } - if (actual_type == OP_number && expected_type == OP_boolean) - *n = expr_allocate_unary (e, OP_NUM_TO_BOOLEAN, *n); break; - case EXPR_STRING: + case VAL_STRING: if (actual_type != OP_string) { msg (SE, _("Type mismatch: expression has %s type, " @@ -499,16 +546,14 @@ match_operator (struct lexer *lexer, const struct operator ops[], size_t op_cnt, const struct operator *op; for (op = ops; op < ops + op_cnt; op++) - { - if (op->token == '-') - lex_negative_to_dash (lexer); - if (lex_match (lexer, op->token)) - { - if (operator != NULL) - *operator = op; - return true; - } - } + if (lex_token (lexer) == op->token) + { + if (op->token != T_NEG_NUM) + lex_get (lexer); + if (operator != NULL) + *operator = op; + return true; + } if (operator != NULL) *operator = NULL; return false; @@ -665,7 +710,7 @@ parse_rel (struct lexer *lexer, struct expression *e) { static const struct operator ops[] = { - { '=', OP_EQ, "numeric equality (`=')" }, + { T_EQUALS, OP_EQ, "numeric equality (`=')" }, { T_EQ, OP_EQ, "numeric equality (`EQ')" }, { T_GE, OP_GE, "numeric greater-than-or-equal-to (`>=')" }, { T_GT, OP_GT, "numeric greater than (`>')" }, @@ -683,7 +728,7 @@ parse_rel (struct lexer *lexer, struct expression *e) { static const struct operator ops[] = { - { '=', OP_EQ_STRING, "string equality (`=')" }, + { T_EQUALS, OP_EQ_STRING, "string equality (`=')" }, { T_EQ, OP_EQ_STRING, "string equality (`EQ')" }, { T_GE, OP_GE_STRING, "string greater-than-or-equal-to (`>=')" }, { T_GT, OP_GT_STRING, "string greater than (`>')" }, @@ -708,8 +753,9 @@ parse_add (struct lexer *lexer, struct expression *e) { static const struct operator ops[] = { - { '+', OP_ADD, "addition (`+')" }, - { '-', OP_SUB, "subtraction (`-')" }, + { T_PLUS, OP_ADD, "addition (`+')" }, + { T_DASH, OP_SUB, "subtraction (`-')" }, + { T_NEG_NUM, OP_ADD, "subtraction (`-')" }, }; return parse_binary_operators (lexer, e, parse_mul (lexer, e), @@ -723,8 +769,8 @@ parse_mul (struct lexer *lexer, struct expression *e) { static const struct operator ops[] = { - { '*', OP_MUL, "multiplication (`*')" }, - { '/', OP_DIV, "division (`/')" }, + { T_ASTERISK, OP_MUL, "multiplication (`*')" }, + { T_SLASH, OP_DIV, "division (`/')" }, }; return parse_binary_operators (lexer, e, parse_neg (lexer, e), @@ -736,7 +782,7 @@ parse_mul (struct lexer *lexer, struct expression *e) static union any_node * parse_neg (struct lexer *lexer, struct expression *e) { - static const struct operator op = { '-', OP_NEG, "negation (`-')" }; + static const struct operator op = { T_DASH, OP_NEG, "negation (`-')" }; return parse_inverting_unary_operator (lexer, e, &op, parse_exp); } @@ -752,8 +798,21 @@ parse_exp (struct lexer *lexer, struct expression *e) "That is, `a**b**c' equals `(a**b)**c', not as `a**(b**c)'. " "To disable this warning, insert parentheses."); - return parse_binary_operators (lexer, e, parse_primary (lexer, e), &op, 1, - parse_primary, chain_warning); + union any_node *lhs, *node; + bool negative = false; + + if (lex_token (lexer) == T_NEG_NUM) + { + lhs = expr_allocate_number (e, -lex_tokval (lexer)); + negative = true; + lex_get (lexer); + } + else + lhs = parse_primary (lexer, e); + + node = parse_binary_operators (lexer, e, lhs, &op, 1, + parse_primary, chain_warning); + return negative ? expr_allocate_unary (e, OP_NEG, node) : node; } /* Parses system variables. */ @@ -773,12 +832,14 @@ parse_sysvar (struct lexer *lexer, struct expression *e) time_t last_proc_time = time_of_last_procedure (e->ds); struct tm *time; char temp_buf[10]; + struct substring s; time = localtime (&last_proc_time); sprintf (temp_buf, "%02d %s %02d", abs (time->tm_mday) % 100, months[abs (time->tm_mon) % 12], abs (time->tm_year) % 100); - return expr_allocate_string_buffer (e, temp_buf, strlen (temp_buf)); + ss_alloc_substring (&s, ss_cstr (temp_buf)); + return expr_allocate_string (e, s); } else if (lex_match_id (lexer, "$TRUE")) return expr_allocate_boolean (e, 1.0); @@ -812,7 +873,7 @@ parse_sysvar (struct lexer *lexer, struct expression *e) return expr_allocate_number (e, settings_get_viewwidth ()); else { - msg (SE, _("Unknown system variable %s."), lex_tokid (lexer)); + msg (SE, _("Unknown system variable %s."), lex_tokcstr (lexer)); return NULL; } } @@ -824,22 +885,22 @@ parse_primary (struct lexer *lexer, struct expression *e) switch (lex_token (lexer)) { case T_ID: - if (lex_look_ahead (lexer) == '(') + if (lex_next_token (lexer, 1) == T_LPAREN) { /* An identifier followed by a left parenthesis may be a vector element reference. If not, it's a function call. */ - if (e->ds != NULL && dict_lookup_vector (dataset_dict (e->ds), lex_tokid (lexer)) != NULL) + if (e->ds != NULL && dict_lookup_vector (dataset_dict (e->ds), lex_tokcstr (lexer)) != NULL) return parse_vector_element (lexer, e); else return parse_function (lexer, e); } - else if (lex_tokid (lexer)[0] == '$') + else if (lex_tokcstr (lexer)[0] == '$') { /* $ at the beginning indicates a system variable. */ return parse_sysvar (lexer, e); } - else if (e->ds != NULL && dict_lookup_var (dataset_dict (e->ds), lex_tokid (lexer))) + else if (e->ds != NULL && dict_lookup_var (dataset_dict (e->ds), lex_tokcstr (lexer))) { /* It looks like a user variable. (It could be a format specifier, but we'll assume @@ -860,7 +921,7 @@ parse_primary (struct lexer *lexer, struct expression *e) return expr_allocate_format (e, &fmt); /* All attempts failed. */ - msg (SE, _("Unknown identifier %s."), lex_tokid (lexer)); + msg (SE, _("Unknown identifier %s."), lex_tokcstr (lexer)); return NULL; } break; @@ -868,34 +929,51 @@ parse_primary (struct lexer *lexer, struct expression *e) case T_POS_NUM: case T_NEG_NUM: { - union any_node *node = expr_allocate_number (e, lex_tokval (lexer) ); + union any_node *node = expr_allocate_number (e, lex_tokval (lexer)); lex_get (lexer); return node; } case T_STRING: { - union any_node *node = expr_allocate_string_buffer ( - e, ds_cstr (lex_tokstr (lexer) ), ds_length (lex_tokstr (lexer) )); + const char *dict_encoding; + union any_node *node; + char *s; + + dict_encoding = (e->ds != NULL + ? dict_get_encoding (dataset_dict (e->ds)) + : "UTF-8"); + s = recode_string_pool (dict_encoding, "UTF-8", lex_tokcstr (lexer), + ss_length (lex_tokss (lexer)), e->expr_pool); + node = expr_allocate_string (e, ss_cstr (s)); + lex_get (lexer); return node; } - case '(': + case T_LPAREN: { - union any_node *node; - lex_get (lexer); - node = parse_or (lexer, e); - if (node != NULL && !lex_match (lexer, ')')) - { - lex_error (lexer, _("expecting `)'")); + /* 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; } default: - lex_error (lexer, _("in expression")); + lex_error (lexer, NULL); return NULL; } } @@ -909,19 +987,19 @@ parse_vector_element (struct lexer *lexer, struct expression *e) /* Find vector, skip token. The caller must already have verified that the current token is the name of a vector. */ - vector = dict_lookup_vector (dataset_dict (e->ds), lex_tokid (lexer)); + vector = dict_lookup_vector (dataset_dict (e->ds), lex_tokcstr (lexer)); assert (vector != NULL); lex_get (lexer); /* Skip left parenthesis token. The caller must have verified that the lookahead is a left parenthesis. */ - assert (lex_token (lexer) == '('); + assert (lex_token (lexer) == T_LPAREN); lex_get (lexer); element = parse_or (lexer, e); if (!type_coercion (e, OP_number, &element, "vector indexing") - || !lex_match (lexer, ')')) + || !lex_match (lexer, T_RPAREN)) return NULL; return expr_allocate_binary (e, (vector_get_type (vector) == VAL_NUMERIC @@ -984,7 +1062,7 @@ compare_names (const char *test, const char *name, bool abbrev_ok) static int compare_strings (const char *test, const char *name, bool abbrev_ok UNUSED) { - return strcasecmp (test, name); + return c_strcasecmp (test, name); } static bool @@ -1024,7 +1102,7 @@ lookup_function (const char *name, } static int -extract_min_valid (char *s) +extract_min_valid (const char *s) { char *p = strrchr (s, '.'); if (p == NULL @@ -1073,23 +1151,21 @@ coerce_function_args (struct expression *e, const struct operation *f, static bool validate_function_args (const struct operation *f, int arg_cnt, int min_valid) { + /* 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_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; - } + assert (array_arg_cnt >= f->array_min_elems); if ((f->flags & OPF_ARRAY_OPERAND) && array_arg_cnt % 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 (arg_cnt % 2 == 0); + msg (SE, _("%s must have an odd number of arguments."), f->prototype); return false; } @@ -1098,26 +1174,19 @@ 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) + if (min_valid > array_arg_cnt) { - 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) - { - 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, arg_cnt, array_arg_cnt, min_valid); return false; } } @@ -1152,7 +1221,7 @@ put_invocation (struct string *s, ds_put_cstr (s, ", "); ds_put_cstr (s, operations[expr_node_returns (args[i])].prototype); } - ds_put_char (s, ')'); + ds_put_byte (s, ')'); } static void @@ -1179,7 +1248,7 @@ no_match (const char *func_name, for (f = first; f < last; f++) ds_put_format (&s, "\n%s", f->prototype); } - ds_put_char (&s, '.'); + ds_put_byte (&s, '.'); msg (SE, "%s", ds_cstr (&s)); @@ -1200,17 +1269,17 @@ parse_function (struct lexer *lexer, struct expression *e) union any_node *n; - ds_init_string (&func_name, lex_tokstr (lexer)); - min_valid = extract_min_valid (ds_cstr (lex_tokstr (lexer))); - if (!lookup_function (ds_cstr (lex_tokstr (lexer)), &first, &last)) + ds_init_substring (&func_name, lex_tokss (lexer)); + min_valid = extract_min_valid (lex_tokcstr (lexer)); + if (!lookup_function (lex_tokcstr (lexer), &first, &last)) { - msg (SE, _("No function or vector named %s."), ds_cstr (lex_tokstr (lexer))); + msg (SE, _("No function or vector named %s."), lex_tokcstr (lexer)); ds_destroy (&func_name); return NULL; } lex_get (lexer); - if (!lex_force_match (lexer, '(')) + if (!lex_force_match (lexer, T_LPAREN)) { ds_destroy (&func_name); return NULL; @@ -1218,11 +1287,11 @@ parse_function (struct lexer *lexer, struct expression *e) args = NULL; arg_cnt = arg_cap = 0; - if (lex_token (lexer) != ')') + if (lex_token (lexer) != T_RPAREN) for (;;) { if (lex_token (lexer) == T_ID - && toupper (lex_look_ahead (lexer)) == 'T') + && lex_next_token (lexer, 1) == T_TO) { const struct variable **vars; size_t var_cnt; @@ -1243,12 +1312,11 @@ parse_function (struct lexer *lexer, struct expression *e) add_arg (&args, &arg_cnt, &arg_cap, arg); } - if (lex_match (lexer, ')')) + if (lex_match (lexer, T_RPAREN)) break; - else if (!lex_match (lexer, ',')) + else if (!lex_match (lexer, T_COMMA)) { - lex_error (lexer, _("expecting `,' or `)' invoking %s function"), - first->name); + lex_error_expecting (lexer, "`,'", "`)'"); goto fail; } } @@ -1270,13 +1338,14 @@ parse_function (struct lexer *lexer, struct expression *e) msg (SW, _("%s is a PSPP extension."), f->prototype); if (f->flags & OPF_UNIMPLEMENTED) { - msg (SE, _("%s is not yet implemented."), f->prototype); + msg (SE, _("%s is not available in this version of PSPP."), + f->prototype); goto fail; } if ((f->flags & OPF_PERM_ONLY) && proc_in_temporary_transformations (e->ds)) { - msg (SE, _("%s may not appear after TEMPORARY."), f->prototype); + msg (SE, _("%s may not appear after %s."), f->prototype, "TEMPORARY"); goto fail; } @@ -1464,18 +1533,6 @@ expr_allocate_vector (struct expression *e, const struct vector *vector) return n; } -union any_node * -expr_allocate_string_buffer (struct expression *e, - const char *string, size_t length) -{ - union any_node *n = pool_alloc (e->expr_pool, sizeof n->string); - n->type = OP_string; - if (length > MAX_STRING) - length = MAX_STRING; - n->string.s = copy_string (e, string, length); - return n; -} - union any_node * expr_allocate_string (struct expression *e, struct substring s) {