X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=src%2Flanguage%2Fcommand.c;h=703ae3ebceaf9c683926f231e31e3e5aa928b1a3;hb=2be9bee9da6a2ce27715e58128569594319abfa2;hp=49ec22fdb8904546cd10d081d89ee31fed15a19a;hpb=92c09e564002d356d20fc1e2e131027ef89f6748;p=pspp-builds.git diff --git a/src/language/command.c b/src/language/command.c index 49ec22fd..703ae3eb 100644 --- a/src/language/command.c +++ b/src/language/command.c @@ -1,54 +1,44 @@ -/* PSPP - computes sample statistics. - Copyright (C) 1997-9, 2000 Free Software Foundation, Inc. +/* PSPP - a program for statistical analysis. + Copyright (C) 1997-9, 2000, 2009, 2010, 2011 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 the Free Software Foundation; either version 2 of the - License, or (at your option) any later version. + 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 + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. - This program is distributed in the hope that it will be useful, but - WITHOUT ANY WARRANTY; without even the implied warranty of - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - General Public License for more details. + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. You should have received a copy of the GNU General Public License - along with this program; if not, write to the Free Software - Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA - 02110-1301, USA. */ + along with this program. If not, see . */ #include -#include +#include "language/command.h" #include #include #include #include -#include - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#if HAVE_SYS_WAIT_H -#include -#endif - -#if HAVE_READLINE -#include -#endif + +#include "data/casereader.h" +#include "data/dataset.h" +#include "data/dictionary.h" +#include "data/settings.h" +#include "data/variable.h" +#include "language/lexer/command-name.h" +#include "language/lexer/lexer.h" +#include "libpspp/assertion.h" +#include "libpspp/compiler.h" +#include "libpspp/i18n.h" +#include "libpspp/message.h" +#include "libpspp/str.h" +#include "output/text-item.h" + +#include "xalloc.h" +#include "xmalloca.h" #include "gettext.h" #define _(msgid) gettext (msgid) @@ -57,7 +47,7 @@ /* Returns true if RESULT is a valid "enum cmd_result", false otherwise. */ static inline bool -cmd_result_is_valid (enum cmd_result result) +cmd_result_is_valid (enum cmd_result result) { return (result == CMD_SUCCESS || result == CMD_EOF || result == CMD_FINISH || (result >= CMD_PRIVATE_FIRST && result <= CMD_PRIVATE_LAST) @@ -68,7 +58,7 @@ cmd_result_is_valid (enum cmd_result result) /* Returns true if RESULT indicates success, false otherwise. */ bool -cmd_result_is_success (enum cmd_result result) +cmd_result_is_success (enum cmd_result result) { assert (cmd_result_is_valid (result)); return result > 0; @@ -77,7 +67,7 @@ cmd_result_is_success (enum cmd_result result) /* Returns true if RESULT indicates failure, false otherwise. */ bool -cmd_result_is_failure (enum cmd_result result) +cmd_result_is_failure (enum cmd_result result) { assert (cmd_result_is_valid (result)); return result < 0; @@ -94,11 +84,10 @@ enum states }; /* Other command requirements. */ -enum flags +enum flags { F_ENHANCED = 0x10, /* Allowed only in enhanced syntax mode. */ F_TESTING = 0x20, /* Allowed only in testing mode. */ - F_KEEP_FINAL_TOKEN = 0x40,/* Don't skip final token in command name. */ F_ABBREV = 0x80 /* Not a candidate for name completion. */ }; @@ -114,7 +103,7 @@ struct command /* Define the command array. */ #define DEF_CMD(STATES, FLAGS, NAME, FUNCTION) {STATES, FLAGS, NAME, FUNCTION}, #define UNIMPL_CMD(NAME, DESCRIPTION) {S_ANY, 0, NAME, NULL}, -static const struct command commands[] = +static const struct command commands[] = { #include "command.def" }; @@ -125,12 +114,12 @@ static const size_t command_cnt = sizeof commands / sizeof *commands; static bool in_correct_state (const struct command *, enum cmd_state); static bool report_state_mismatch (const struct command *, enum cmd_state); -static const struct command *find_command (const char *name); -static void set_completion_state (enum cmd_state); +static void set_completion_state (enum cmd_state); /* Command parser. */ -static const struct command *parse_command_name (struct lexer *lexer); +static const struct command *parse_command_name (struct lexer *, + int *n_tokens); static enum cmd_result do_parse_command (struct lexer *, struct dataset *, enum cmd_state); /* Parses an entire command, from command name to terminating @@ -142,11 +131,7 @@ cmd_parse_in_state (struct lexer *lexer, struct dataset *ds, { int result; - som_new_series (); - result = do_parse_command (lexer, ds, state); - if (cmd_result_is_failure (result)) - lex_discard_rest_of_command (lexer); assert (!proc_is_open (ds)); unset_cmd_algorithm (); @@ -171,413 +156,193 @@ cmd_parse (struct lexer *lexer, struct dataset *ds) /* Parses an entire command, from command name to terminating dot. */ static enum cmd_result -do_parse_command (struct lexer *lexer, struct dataset *ds, enum cmd_state state) +do_parse_command (struct lexer *lexer, + struct dataset *ds, enum cmd_state state) { - const struct command *command; + const struct command *command = NULL; enum cmd_result result; + bool opened = false; + int n_tokens; /* Read the command's first token. */ - prompt_set_style (PROMPT_FIRST); set_completion_state (state); - lex_get (lexer); if (lex_token (lexer) == T_STOP) - return CMD_EOF; - else if (lex_token (lexer) == '.') + { + result = CMD_EOF; + goto finish; + } + else if (lex_token (lexer) == T_ENDCMD) { /* Null commands can result from extra empty lines. */ - return CMD_SUCCESS; + result = CMD_SUCCESS; + goto finish; } - prompt_set_style (PROMPT_LATER); /* Parse the command name. */ - command = parse_command_name (lexer); + command = parse_command_name (lexer, &n_tokens); if (command == NULL) - return CMD_FAILURE; - else if (command->function == NULL) { - msg (SE, _("%s is unimplemented."), command->name); - return CMD_NOT_IMPLEMENTED; + result = CMD_FAILURE; + goto finish; + } + text_item_submit (text_item_create (TEXT_ITEM_COMMAND_OPEN, command->name)); + opened = true; + + if (command->function == NULL) + { + msg (SE, _("%s is not yet implemented."), command->name); + result = CMD_NOT_IMPLEMENTED; } - else if ((command->flags & F_TESTING) && !get_testing_mode ()) + else if ((command->flags & F_TESTING) && !settings_get_testing_mode ()) { msg (SE, _("%s may be used only in testing mode."), command->name); - return CMD_FAILURE; + result = CMD_FAILURE; } - else if ((command->flags & F_ENHANCED) && get_syntax () != ENHANCED) + else if ((command->flags & F_ENHANCED) && settings_get_syntax () != ENHANCED) { msg (SE, _("%s may be used only in enhanced syntax mode."), command->name); - return CMD_FAILURE; + result = CMD_FAILURE; } - else if (!in_correct_state (command, state)) + else if (!in_correct_state (command, state)) { report_state_mismatch (command, state); - return CMD_FAILURE; - } - - /* Execute command. */ - msg_set_command_name (command->name); - tab_set_command_name (command->name); - result = command->function (lexer, ds); - tab_set_command_name (NULL); - msg_set_command_name (NULL); - - assert (cmd_result_is_valid (result)); - return result; -} - -static size_t -match_strings (const char *a, size_t a_len, - const char *b, size_t b_len) -{ - size_t match_len = 0; - - while (a_len > 0 && b_len > 0) - { - /* Mismatch always returns zero. */ - if (toupper ((unsigned char) *a++) != toupper ((unsigned char) *b++)) - return 0; - - /* Advance. */ - a_len--; - b_len--; - match_len++; + result = CMD_FAILURE; } - - return match_len; -} - -/* Returns the first character in the first word in STRING, - storing the word's length in *WORD_LEN. If no words remain, - returns a null pointer and stores 0 in *WORD_LEN. Words are - sequences of alphanumeric characters or single - non-alphanumeric characters. Words are delimited by - spaces. */ -static const char * -find_word (const char *string, size_t *word_len) -{ - /* Skip whitespace and asterisks. */ - while (isspace ((unsigned char) *string)) - string++; - - /* End of string? */ - if (*string == '\0') + else { - *word_len = 0; - return NULL; - } + /* Execute command. */ + int i; - /* Special one-character word? */ - if (!isalnum ((unsigned char) *string)) - { - *word_len = 1; - return string; + for (i = 0; i < n_tokens; i++) + lex_get (lexer); + result = command->function (lexer, ds); } - /* Alphanumeric word. */ - *word_len = 1; - while (isalnum ((unsigned char) string[*word_len])) - (*word_len)++; - - return string; -} - -/* Returns true if strings A and B can be confused based on - their first three letters. */ -static bool -conflicting_3char_prefixes (const char *a, const char *b) -{ - size_t aw_len, bw_len; - const char *aw, *bw; - - aw = find_word (a, &aw_len); - bw = find_word (b, &bw_len); - assert (aw != NULL && bw != NULL); - - /* Words that are the same don't conflict. */ - if (aw_len == bw_len && !buf_compare_case (aw, bw, aw_len)) - return false; - - /* Words that are otherwise the same in the first three letters - do conflict. */ - return ((aw_len > 3 && bw_len > 3) - || (aw_len == 3 && bw_len > 3) - || (bw_len == 3 && aw_len > 3)) && !buf_compare_case (aw, bw, 3); -} + assert (cmd_result_is_valid (result)); -/* Returns true if CMD can be confused with another command - based on the first three letters of its first word. */ -static bool -conflicting_3char_prefix_command (const struct command *cmd) -{ - assert (cmd >= commands && cmd < commands + command_cnt); +finish: + if (cmd_result_is_failure (result)) + lex_interactive_reset (lexer); + else if (result == CMD_SUCCESS) + result = lex_end_of_command (lexer); - return ((cmd > commands - && conflicting_3char_prefixes (cmd[-1].name, cmd[0].name)) - || (cmd < commands + command_cnt - && conflicting_3char_prefixes (cmd[0].name, cmd[1].name))); -} + lex_discard_rest_of_command (lexer); + while (lex_token (lexer) == T_ENDCMD) + lex_get (lexer); -/* Ways that a set of words can match a command name. */ -enum command_match - { - MISMATCH, /* Not a match. */ - PARTIAL_MATCH, /* The words begin the command name. */ - COMPLETE_MATCH /* The words are the command name. */ - }; + if (opened) + text_item_submit (text_item_create (TEXT_ITEM_COMMAND_CLOSE, + command->name)); -/* Figures out how well the WORD_CNT words in WORDS match CMD, - and returns the appropriate enum value. If WORDS are a - partial match for CMD and the next word in CMD is a dash, then - *DASH_POSSIBLE is set to 1 if DASH_POSSIBLE is non-null; - otherwise, *DASH_POSSIBLE is unchanged. */ -static enum command_match -cmd_match_words (const struct command *cmd, - char *const words[], size_t word_cnt, - int *dash_possible) -{ - const char *word; - size_t word_len; - size_t word_idx; - - for (word = find_word (cmd->name, &word_len), word_idx = 0; - word != NULL && word_idx < word_cnt; - word = find_word (word + word_len, &word_len), word_idx++) - if (word_len != strlen (words[word_idx]) - || buf_compare_case (word, words[word_idx], word_len)) - { - size_t match_chars = match_strings (word, word_len, - words[word_idx], - strlen (words[word_idx])); - if (match_chars == 0) - { - /* Mismatch. */ - return MISMATCH; - } - else if (match_chars == 1 || match_chars == 2) - { - /* One- and two-character abbreviations are not - acceptable. */ - return MISMATCH; - } - else if (match_chars == 3) - { - /* Three-character abbreviations are acceptable - in the first word of a command if there are - no name conflicts. They are always - acceptable after the first word. */ - if (word_idx == 0 && conflicting_3char_prefix_command (cmd)) - return MISMATCH; - } - else /* match_chars > 3 */ - { - /* Four-character and longer abbreviations are - always acceptable. */ - } - } - - if (word == NULL && word_idx == word_cnt) - { - /* cmd->name = "FOO BAR", words[] = {"FOO", "BAR"}. */ - return COMPLETE_MATCH; - } - else if (word == NULL) - { - /* cmd->name = "FOO BAR", words[] = {"FOO", "BAR", "BAZ"}. */ - return MISMATCH; - } - else - { - /* cmd->name = "FOO BAR BAZ", words[] = {"FOO", "BAR"}. */ - if (word[0] == '-' && dash_possible != NULL) - *dash_possible = 1; - return PARTIAL_MATCH; - } + return result; } -/* Returns the number of commands for which the WORD_CNT words in - WORDS are a partial or complete match. If some partial match - has a dash as the next word, then *DASH_POSSIBLE is set to 1, - otherwise it is set to 0. */ static int -count_matching_commands (char *const words[], size_t word_cnt, - int *dash_possible) +find_best_match (struct substring s, const struct command **matchp) { const struct command *cmd; - int cmd_match_count; + struct command_matcher cm; + int missing_words; - cmd_match_count = 0; - *dash_possible = 0; - for (cmd = commands; cmd < commands + command_cnt; cmd++) - if (cmd_match_words (cmd, words, word_cnt, dash_possible) != MISMATCH) - cmd_match_count++; + command_matcher_init (&cm, s); + for (cmd = commands; cmd < &commands[command_cnt]; cmd++) + command_matcher_add (&cm, ss_cstr (cmd->name), CONST_CAST (void *, cmd)); - return cmd_match_count; -} + *matchp = command_matcher_get_match (&cm); + missing_words = command_matcher_get_missing_words (&cm); -/* Returns the command for which the WORD_CNT words in WORDS are - a complete match. Returns a null pointer if no such command - exists. */ -static const struct command * -get_complete_match (char *const words[], size_t word_cnt) -{ - const struct command *cmd; - - for (cmd = commands; cmd < commands + command_cnt; cmd++) - if (cmd_match_words (cmd, words, word_cnt, NULL) == COMPLETE_MATCH) - return cmd; - - return NULL; -} + command_matcher_destroy (&cm); -/* Returns the command with the given exact NAME. - Aborts if no such command exists. */ -static const struct command * -find_command (const char *name) -{ - const struct command *cmd; - - for (cmd = commands; cmd < commands + command_cnt; cmd++) - if (!strcmp (cmd->name, name)) - return cmd; - NOT_REACHED (); + return missing_words; } -/* Frees the WORD_CNT words in WORDS. */ -static void -free_words (char *words[], size_t word_cnt) +static bool +parse_command_word (struct lexer *lexer, struct string *s, int n) { - size_t idx; - - for (idx = 0; idx < word_cnt; idx++) - free (words[idx]); -} + bool need_space = ds_last (s) != EOF && ds_last (s) != '-'; -/* Flags an error that the command whose name is given by the - WORD_CNT words in WORDS is unknown. */ -static void -unknown_command_error (struct lexer *lexer, char *const words[], size_t word_cnt) -{ - if (word_cnt == 0) - lex_error (lexer, _("expecting command name")); - else + switch (lex_next_token (lexer, n)) { - struct string s; - size_t i; + case T_DASH: + ds_put_byte (s, '-'); + return true; + + case T_ID: + if (need_space) + ds_put_byte (s, ' '); + ds_put_cstr (s, lex_next_tokcstr (lexer, n)); + return true; - ds_init_empty (&s); - for (i = 0; i < word_cnt; i++) + case T_POS_NUM: + if (lex_next_is_integer (lexer, n)) { - if (i != 0) - ds_put_char (&s, ' '); - ds_put_cstr (&s, words[i]); + int integer = lex_next_integer (lexer, n); + if (integer >= 0) + { + if (need_space) + ds_put_byte (s, ' '); + ds_put_format (s, "%ld", lex_next_integer (lexer, n)); + return true; + } } + return false; - msg (SE, _("Unknown command %s."), ds_cstr (&s)); - - ds_destroy (&s); + default: + return false; } } -/* Parse the command name and return a pointer to the corresponding - struct command if successful. - If not successful, return a null pointer. */ +/* Parses the command name. On success returns a pointer to the corresponding + struct command and stores the number of tokens in the command name into + *N_TOKENS. On failure, returns a null pointer and stores the number of + tokens required to determine that no command name was present into + *N_TOKENS. */ static const struct command * -parse_command_name (struct lexer *lexer) +parse_command_name (struct lexer *lexer, int *n_tokens) { - char *words[16]; - int word_cnt; - int complete_word_cnt; - int dash_possible; - - if (lex_token (lexer) == T_EXP || - lex_token (lexer) == '*' || lex_token (lexer) == '[') - return find_command ("COMMENT"); - - dash_possible = 0; - word_cnt = complete_word_cnt = 0; - while (lex_token (lexer) == T_ID || (dash_possible && lex_token (lexer) == '-')) + const struct command *command; + int missing_words; + struct string s; + int word; + + command = NULL; + missing_words = 0; + ds_init_empty (&s); + word = 0; + while (parse_command_word (lexer, &s, word)) { - int cmd_match_cnt; - - assert (word_cnt < sizeof words / sizeof *words); - if (lex_token (lexer) == T_ID) - { - words[word_cnt] = ds_xstrdup (lex_tokstr (lexer)); - str_uppercase (words[word_cnt]); - } - else if (lex_token (lexer) == '-') - words[word_cnt] = xstrdup ("-"); - word_cnt++; - - cmd_match_cnt = count_matching_commands (words, word_cnt, - &dash_possible); - if (cmd_match_cnt == 0) + missing_words = find_best_match (ds_ss (&s), &command); + if (missing_words <= 0) break; - else if (cmd_match_cnt == 1) - { - const struct command *command = get_complete_match (words, word_cnt); - if (command != NULL) - { - if (!(command->flags & F_KEEP_FINAL_TOKEN)) - lex_get (lexer); - free_words (words, word_cnt); - return command; - } - } - else /* cmd_match_cnt > 1 */ - { - /* Do we have a complete command name so far? */ - if (get_complete_match (words, word_cnt) != NULL) - complete_word_cnt = word_cnt; - } - lex_get (lexer); + word++; } - /* If we saw a complete command name earlier, drop back to - it. */ - if (complete_word_cnt) + if (command == NULL && missing_words > 0) { - int pushback_word_cnt; - const struct command *command; - - /* Get the command. */ - command = get_complete_match (words, complete_word_cnt); - assert (command != NULL); - - /* Figure out how many words we want to keep. - We normally want to swallow the entire command. */ - pushback_word_cnt = complete_word_cnt + 1; - if (command->flags & F_KEEP_FINAL_TOKEN) - pushback_word_cnt--; - - /* FIXME: We only support one-token pushback. */ - assert (pushback_word_cnt + 1 >= word_cnt); - - while (word_cnt > pushback_word_cnt) - { - word_cnt--; - if (strcmp (words[word_cnt], "-")) - lex_put_back_id (lexer, words[word_cnt]); - else - lex_put_back (lexer, '-'); - free (words[word_cnt]); - } + ds_put_cstr (&s, " ."); + missing_words = find_best_match (ds_ss (&s), &command); + ds_truncate (&s, ds_length (&s) - 2); + } - free_words (words, word_cnt); - return command; + if (command == NULL) + { + if (ds_is_empty (&s)) + lex_error (lexer, _("expecting command name")); + else + msg (SE, _("Unknown command `%s'."), ds_cstr (&s)); } - /* We didn't get a valid command name. */ - unknown_command_error (lexer, words, word_cnt); - free_words (words, word_cnt); - return NULL; + ds_destroy (&s); + + *n_tokens = (word + 1) + missing_words; + return command; } /* Returns true if COMMAND is allowed in STATE, false otherwise. */ static bool -in_correct_state (const struct command *command, enum cmd_state state) +in_correct_state (const struct command *command, enum cmd_state state) { return ((state == CMD_STATE_INITIAL && command->states & S_INITIAL) || (state == CMD_STATE_DATA && command->states & S_DATA) @@ -594,37 +359,78 @@ report_state_mismatch (const struct command *command, enum cmd_state state) assert (!in_correct_state (command, state)); if (state == CMD_STATE_INITIAL || state == CMD_STATE_DATA) { - const char *allowed[3]; - int allowed_cnt; - char *s; - - allowed_cnt = 0; - if (command->states & S_INITIAL) - allowed[allowed_cnt++] = _("before the active file has been defined"); - else if (command->states & S_DATA) - allowed[allowed_cnt++] = _("after the active file has been defined"); - if (command->states & S_INPUT_PROGRAM) - allowed[allowed_cnt++] = _("inside INPUT PROGRAM"); - if (command->states & S_FILE_TYPE) - allowed[allowed_cnt++] = _("inside FILE TYPE"); - - if (allowed_cnt == 1) - s = xstrdup (allowed[0]); - else if (allowed_cnt == 2) - s = xasprintf (_("%s or %s"), allowed[0], allowed[1]); - else if (allowed_cnt == 3) - s = xasprintf (_("%s, %s, or %s"), allowed[0], allowed[1], allowed[2]); - else - NOT_REACHED (); - - msg (SE, _("%s is allowed only %s."), command->name, s); - - free (s); + switch (command->states) + { + /* One allowed state. */ + case S_INITIAL: + msg (SE, _("%s is allowed only before the active file has " + "been defined."), command->name); + break; + case S_DATA: + msg (SE, _("%s is allowed only after the active file has " + "been defined."), command->name); + break; + case S_INPUT_PROGRAM: + msg (SE, _("%s is allowed only inside INPUT PROGRAM."), + command->name); + break; + case S_FILE_TYPE: + msg (SE, _("%s is allowed only inside FILE TYPE."), command->name); + break; + + /* Two allowed states. */ + case S_INITIAL | S_DATA: + NOT_REACHED (); + case S_INITIAL | S_INPUT_PROGRAM: + msg (SE, _("%s is allowed only before the active file has " + "been defined or inside INPUT PROGRAM."), command->name); + break; + case S_INITIAL | S_FILE_TYPE: + msg (SE, _("%s is allowed only before the active file has " + "been defined or inside FILE TYPE."), command->name); + break; + case S_DATA | S_INPUT_PROGRAM: + msg (SE, _("%s is allowed only after the active file has " + "been defined or inside INPUT PROGRAM."), command->name); + break; + case S_DATA | S_FILE_TYPE: + msg (SE, _("%s is allowed only after the active file has " + "been defined or inside FILE TYPE."), command->name); + break; + case S_INPUT_PROGRAM | S_FILE_TYPE: + msg (SE, _("%s is allowed only inside INPUT PROGRAM " + "or inside FILE TYPE."), command->name); + break; + + /* Three allowed states. */ + case S_DATA | S_INPUT_PROGRAM | S_FILE_TYPE: + msg (SE, _("%s is allowed only after the active file has " + "been defined, inside INPUT PROGRAM, or inside " + "FILE TYPE."), command->name); + break; + case S_INITIAL | S_INPUT_PROGRAM | S_FILE_TYPE: + msg (SE, _("%s is allowed only before the active file has " + "been defined, inside INPUT PROGRAM, or inside " + "FILE TYPE."), command->name); + break; + case S_INITIAL | S_DATA | S_FILE_TYPE: + NOT_REACHED (); + case S_INITIAL | S_DATA | S_INPUT_PROGRAM: + NOT_REACHED (); + + /* Four allowed states. */ + case S_INITIAL | S_DATA | S_INPUT_PROGRAM | S_FILE_TYPE: + NOT_REACHED (); + + default: + NOT_REACHED (); + } } else if (state == CMD_STATE_INPUT_PROGRAM) - msg (SE, _("%s is not allowed inside INPUT PROGRAM."), command->name); + msg (SE, _("%s is not allowed inside %s."), + command->name, "INPUT PROGRAM" ); else if (state == CMD_STATE_FILE_TYPE) - msg (SE, _("%s is not allowed inside FILE TYPE."), command->name); + msg (SE, _("%s is not allowed inside %s."), command->name, "FILE TYPE"); return false; } @@ -634,7 +440,7 @@ report_state_mismatch (const struct command *command, enum cmd_state state) static enum cmd_state completion_state = CMD_STATE_INITIAL; static void -set_completion_state (enum cmd_state state) +set_completion_state (enum cmd_state state) { completion_state = state; } @@ -649,10 +455,10 @@ cmd_complete (const char *prefix, const struct command **cmd) if (*cmd == NULL) *cmd = commands; - for (; *cmd < commands + command_cnt; (*cmd)++) + for (; *cmd < commands + command_cnt; (*cmd)++) if (!memcasecmp ((*cmd)->name, prefix, strlen (prefix)) - && (!((*cmd)->flags & F_TESTING) || get_testing_mode ()) - && (!((*cmd)->flags & F_ENHANCED) || get_syntax () == ENHANCED) + && (!((*cmd)->flags & F_TESTING) || settings_get_testing_mode ()) + && (!((*cmd)->flags & F_ENHANCED) || settings_get_syntax () == ENHANCED) && !((*cmd)->flags & F_ABBREV) && ((*cmd)->function != NULL) && in_correct_state (*cmd, completion_state)) @@ -684,176 +490,57 @@ cmd_n_of_cases (struct lexer *lexer, struct dataset *ds) if (!lex_match_id (lexer, "ESTIMATED")) dict_set_case_limit (dataset_dict (ds), x); - return lex_end_of_command (lexer); + return CMD_SUCCESS; } /* Parses, performs the EXECUTE procedure. */ int -cmd_execute (struct lexer *lexer, struct dataset *ds) +cmd_execute (struct lexer *lexer UNUSED, struct dataset *ds) { bool ok = casereader_destroy (proc_open (ds)); if (!proc_commit (ds) || !ok) return CMD_CASCADING_FAILURE; - return lex_end_of_command (lexer); + return CMD_SUCCESS; } /* Parses, performs the ERASE command. */ int cmd_erase (struct lexer *lexer, struct dataset *ds UNUSED) { - if (get_safer_mode ()) - { - msg (SE, _("This command not allowed when the SAFER option is set.")); - return CMD_FAILURE; - } - + char *filename; + int retval; + + if (settings_get_safer_mode ()) + { + msg (SE, _("This command not allowed when the SAFER option is set.")); + return CMD_FAILURE; + } + if (!lex_force_match_id (lexer, "FILE")) return CMD_FAILURE; - lex_match (lexer, '='); + lex_match (lexer, T_EQUALS); if (!lex_force_string (lexer)) return CMD_FAILURE; - if (remove (ds_cstr (lex_tokstr (lexer))) == -1) + filename = utf8_to_filename (lex_tokcstr (lexer)); + retval = remove (filename); + free (filename); + + if (retval == -1) { msg (SW, _("Error removing `%s': %s."), - ds_cstr (lex_tokstr (lexer)), strerror (errno)); + lex_tokcstr (lexer), strerror (errno)); return CMD_FAILURE; } + lex_get (lexer); return CMD_SUCCESS; } -#if HAVE_FORK && HAVE_EXECL -/* Spawn an interactive shell process. */ -static bool -shell (void) -{ - int pid; - - pid = fork (); - switch (pid) - { - case 0: - { - const char *shell_fn; - char *shell_process; - - { - int i; - - for (i = 3; i < 20; i++) - close (i); - } - - shell_fn = getenv ("SHELL"); - if (shell_fn == NULL) - shell_fn = "/bin/sh"; - - { - const char *cp = strrchr (shell_fn, '/'); - cp = cp ? &cp[1] : shell_fn; - shell_process = local_alloc (strlen (cp) + 8); - strcpy (shell_process, "-"); - strcat (shell_process, cp); - if (strcmp (cp, "sh")) - shell_process[0] = '+'; - } - - execl (shell_fn, shell_process, NULL); - - _exit (1); - } - - case -1: - msg (SE, _("Couldn't fork: %s."), strerror (errno)); - return false; - - default: - assert (pid > 0); - while (wait (NULL) != pid) - ; - return true; - } -} -#else /* !(HAVE_FORK && HAVE_EXECL) */ -/* Don't know how to spawn an interactive shell. */ -static bool -shell (void) -{ - msg (SE, _("Interactive shell not supported on this platform.")); - return false; -} -#endif - -/* Executes the specified COMMAND in a subshell. Returns true if - successful, false otherwise. */ -static bool -run_command (const char *command) -{ - if (system (NULL) == 0) - { - msg (SE, _("Command shell not supported on this platform.")); - return false; - } - - /* Execute the command. */ - if (system (command) == -1) - msg (SE, _("Error executing command: %s."), strerror (errno)); - - return true; -} - -/* Parses, performs the HOST command. */ -int -cmd_host (struct lexer *lexer, struct dataset *ds UNUSED) -{ - int look_ahead; - - if (get_safer_mode ()) - { - msg (SE, _("This command not allowed when the SAFER option is set.")); - return CMD_FAILURE; - } - - look_ahead = lex_look_ahead (lexer); - if (look_ahead == '.') - { - lex_get (lexer); - return shell () ? CMD_SUCCESS : CMD_FAILURE; - } - else if (look_ahead == '\'' || look_ahead == '"') - { - bool ok; - - lex_get (lexer); - if (!lex_force_string (lexer)) - NOT_REACHED (); - ok = run_command (ds_cstr (lex_tokstr (lexer))); - - lex_get (lexer); - return ok ? lex_end_of_command (lexer) : CMD_FAILURE; - } - else - { - bool ok = run_command (lex_rest_of_line (lexer)); - lex_discard_line (lexer); - return ok ? CMD_SUCCESS : CMD_FAILURE; - } -} - /* Parses, performs the NEW FILE command. */ int -cmd_new_file (struct lexer *lexer, struct dataset *ds) +cmd_new_file (struct lexer *lexer UNUSED, struct dataset *ds) { proc_discard_active_file (ds); - - return lex_end_of_command (lexer); -} - -/* Parses a comment. */ -int -cmd_comment (struct lexer *lexer, struct dataset *ds UNUSED) -{ - lex_skip_comment (lexer); return CMD_SUCCESS; }