+/* PSPP - a program for statistical analysis.
+ Copyright (C) 2021 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 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.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>. */
+
+#include <config.h>
+
+#include "language/lexer/macro.h"
+
+#include <errno.h>
+#include <limits.h>
+#include <stdlib.h>
+
+#include "data/settings.h"
+#include "language/lexer/lexer.h"
+#include "language/lexer/segment.h"
+#include "language/lexer/scan.h"
+#include "libpspp/assertion.h"
+#include "libpspp/cast.h"
+#include "libpspp/i18n.h"
+#include "libpspp/message.h"
+#include "libpspp/str.h"
+#include "libpspp/string-array.h"
+#include "libpspp/string-map.h"
+#include "libpspp/stringi-set.h"
+
+#include "gl/c-ctype.h"
+#include "gl/ftoastr.h"
+
+#include "gettext.h"
+#define _(msgid) gettext (msgid)
+
+/* An entry in the stack of macros and macro directives being expanded. The
+ stack is maintained as a linked list. Entries are not dynamically allocated
+ but on the program stack. */
+struct macro_expansion_stack
+ {
+ /* Points to an outer stack entry, or NULL if this is the outermost. */
+ const struct macro_expansion_stack *next;
+
+ /* A macro name or !IF, !DO, etc. */
+ const char *name;
+
+ /* Location of the macro definition, if available. */
+ const char *file_name;
+ int first_line;
+ int last_line;
+ };
+
+/* Reports an error during macro expansion. STACK is the stack for reporting
+ the location of the error, MT is the optional token at which the error was
+ detected, and FORMAT along with the varargs is the message to report. */
+static void PRINTF_FORMAT (3, 4)
+macro_error (const struct macro_expansion_stack *stack,
+ const struct macro_token *mt,
+ const char *format, ...)
+{
+ struct msg_stack **ms = NULL;
+ size_t allocated_ms = 0;
+ size_t n_ms = 0;
+
+ for (const struct macro_expansion_stack *p = stack; p; p = p->next)
+ {
+ if (n_ms >= allocated_ms)
+ ms = x2nrealloc (ms, &allocated_ms, sizeof *ms);
+
+ /* TRANSLATORS: These strings are used for explaining the context of an
+ error. The "While expanding" message appears first, followed by zero
+ or more of the "inside expansion" messages. `innermost',
+ `next_inner`, etc., are names of macros, and `foobar' is a piece of
+ PSPP syntax:
+
+ foo.sps:12: At `foobar' in the expansion of 'innermost',
+ foo.sps:23: inside the expansion of 'next_inner',
+ foo.sps:34: inside the expansion of 'next_inner2',
+ foo.sps:45: inside the expansion of 'outermost',
+ foo.sps:76: This is the actual error message. */
+ char *description;
+ if (p == stack)
+ {
+ if (mt && mt->representation.length)
+ {
+ char syntax[64];
+ str_ellipsize (mt->representation, syntax, sizeof syntax);
+ description = xasprintf (_("At `%s' in the expansion of `%s',"),
+ syntax, p->name);
+ }
+ else
+ description = xasprintf (_("In the expansion of `%s',"), p->name);
+ }
+ else
+ description = xasprintf (_("inside the expansion of `%s',"), p->name);
+
+ ms[n_ms] = xmalloc (sizeof *ms[n_ms]);
+ *ms[n_ms] = (struct msg_stack) {
+ .location = {
+ .file_name = xstrdup_if_nonnull (p->file_name),
+ .first_line = p->first_line,
+ .last_line = p->last_line,
+ },
+ .description = description,
+ };
+ n_ms++;
+ }
+
+ va_list args;
+ va_start (args, format);
+ char *s = xvasprintf (format, args);
+ va_end (args);
+
+ struct msg *m = xmalloc (sizeof *m);
+ *m = (struct msg) {
+ .category = MSG_C_SYNTAX,
+ .severity = MSG_S_ERROR,
+ .stack = ms,
+ .n_stack = n_ms,
+ .text = s,
+ };
+ msg_emit (m);
+}
+
+void
+macro_token_copy (struct macro_token *dst, const struct macro_token *src)
+{
+ token_copy (&dst->token, &src->token);
+ ss_alloc_substring (&dst->representation, src->representation);
+}
+
+void
+macro_token_uninit (struct macro_token *mt)
+{
+ token_uninit (&mt->token);
+ ss_dealloc (&mt->representation);
+}
+
+void
+macro_token_to_representation (struct macro_token *mt, struct string *s)
+{
+ ds_put_substring (s, mt->representation);
+}
+bool
+is_macro_keyword (struct substring s)
+{
+ static struct stringi_set keywords = STRINGI_SET_INITIALIZER (keywords);
+ if (stringi_set_is_empty (&keywords))
+ {
+ static const char *kws[] = {
+ "BREAK",
+ "CHAREND",
+ "CMDEND",
+ "DEFAULT",
+ "DO",
+ "DOEND",
+ "ELSE",
+ "ENCLOSE",
+ "ENDDEFINE",
+ "IF",
+ "IFEND",
+ "IN",
+ "LET",
+ "NOEXPAND",
+ "OFFEXPAND",
+ "ONEXPAND",
+ "POSITIONAL",
+ "THEN",
+ "TOKENS",
+ };
+ for (size_t i = 0; i < sizeof kws / sizeof *kws; i++)
+ stringi_set_insert (&keywords, kws[i]);
+ }
+
+ ss_ltrim (&s, ss_cstr ("!"));
+ return stringi_set_contains_len (&keywords, s.string, s.length);
+}
+\f
+void
+macro_tokens_copy (struct macro_tokens *dst, const struct macro_tokens *src)
+{
+ *dst = (struct macro_tokens) {
+ .mts = xmalloc (src->n * sizeof *dst->mts),
+ .n = src->n,
+ .allocated = src->n,
+ };
+ for (size_t i = 0; i < src->n; i++)
+ macro_token_copy (&dst->mts[i], &src->mts[i]);
+}
+
+void
+macro_tokens_uninit (struct macro_tokens *mts)
+{
+ for (size_t i = 0; i < mts->n; i++)
+ macro_token_uninit (&mts->mts[i]);
+ free (mts->mts);
+}
+
+struct macro_token *
+macro_tokens_add_uninit (struct macro_tokens *mts)
+{
+ if (mts->n >= mts->allocated)
+ mts->mts = x2nrealloc (mts->mts, &mts->allocated, sizeof *mts->mts);
+ return &mts->mts[mts->n++];
+}
+
+void
+macro_tokens_add (struct macro_tokens *mts, const struct macro_token *mt)
+{
+ macro_token_copy (macro_tokens_add_uninit (mts), mt);
+}
+
+/* Tokenizes SRC according to MODE and appends the tokens to MTS. Uses STACK,
+ if nonull, for error reporting. */
+static void
+macro_tokens_from_string__ (struct macro_tokens *mts, const struct substring src,
+ enum segmenter_mode mode,
+ const struct macro_expansion_stack *stack)
+{
+ struct state
+ {
+ struct segmenter segmenter;
+ struct substring body;
+ };
+
+ struct state state = {
+ .segmenter = segmenter_init (mode, true),
+ .body = src,
+ };
+ struct state saved = state;
+
+ while (state.body.length > 0)
+ {
+ struct macro_token mt = {
+ .token = { .type = T_STOP },
+ .representation = { .string = state.body.string },
+ };
+ struct token *token = &mt.token;
+
+ struct scanner scanner;
+ scanner_init (&scanner, token);
+
+ for (;;)
+ {
+ enum segment_type type;
+ int seg_len = segmenter_push (&state.segmenter, state.body.string,
+ state.body.length, true, &type);
+ assert (seg_len >= 0);
+
+ struct substring segment = ss_head (state.body, seg_len);
+ ss_advance (&state.body, seg_len);
+
+ enum scan_result result = scanner_push (&scanner, type, segment, token);
+ if (result == SCAN_SAVE)
+ saved = state;
+ else if (result == SCAN_BACK)
+ {
+ state = saved;
+ break;
+ }
+ else if (result == SCAN_DONE)
+ break;
+ }
+
+ /* We have a token in 'token'. */
+ mt.representation.length = state.body.string - mt.representation.string;
+ if (is_scan_type (token->type))
+ {
+ if (token->type != SCAN_SKIP)
+ {
+ char *s = scan_token_to_error (token);
+ if (stack)
+ {
+ mt.token.type = T_STRING;
+ macro_error (stack, &mt, "%s", s);
+ }
+ else
+ msg (SE, "%s", s);
+ free (s);
+ }
+ }
+ else
+ macro_tokens_add (mts, &mt);
+ token_uninit (token);
+ }
+}
+
+/* Tokenizes SRC according to MODE and appends the tokens to MTS. */
+void
+macro_tokens_from_string (struct macro_tokens *mts, const struct substring src,
+ enum segmenter_mode mode)
+{
+ macro_tokens_from_string__ (mts, src, mode, NULL);
+}
+
+void
+macro_tokens_print (const struct macro_tokens *mts, FILE *stream)
+{
+ for (size_t i = 0; i < mts->n; i++)
+ token_print (&mts->mts[i].token, stream);
+}
+
+enum token_class
+ {
+ TC_ENDCMD, /* No space before or after (new-line after). */
+ TC_BINOP, /* Space on both sides. */
+ TC_COMMA, /* Space afterward. */
+ TC_ID, /* Don't need spaces except sequentially. */
+ TC_PUNCT, /* Don't need spaces except sequentially. */
+ };
+
+static bool
+needs_space (enum token_class prev, enum token_class next)
+{
+ /* Don't need a space before or after the end of a command.
+ (A new-line is needed afterward as a special case.) */
+ if (prev == TC_ENDCMD || next == TC_ENDCMD)
+ return false;
+
+ /* Binary operators always have a space on both sides. */
+ if (prev == TC_BINOP || next == TC_BINOP)
+ return true;
+
+ /* A comma always has a space afterward. */
+ if (prev == TC_COMMA)
+ return true;
+
+ /* Otherwise, PREV is TC_ID or TC_PUNCT, which only need a space if there are
+ two or them in a row. */
+ return prev == next;
+}
+
+static enum token_class
+classify_token (enum token_type type)
+{
+ switch (type)
+ {
+ case T_ID:
+ case T_MACRO_ID:
+ case T_POS_NUM:
+ case T_NEG_NUM:
+ case T_STRING:
+ return TC_ID;
+
+ case T_STOP:
+ return TC_PUNCT;
+
+ case T_ENDCMD:
+ return TC_ENDCMD;
+
+ case T_LPAREN:
+ case T_RPAREN:
+ case T_LBRACK:
+ case T_RBRACK:
+ return TC_PUNCT;
+
+ case T_PLUS:
+ case T_DASH:
+ case T_ASTERISK:
+ case T_SLASH:
+ case T_EQUALS:
+ case T_AND:
+ case T_OR:
+ case T_NOT:
+ case T_EQ:
+ case T_GE:
+ case T_GT:
+ case T_LE:
+ case T_LT:
+ case T_NE:
+ case T_ALL:
+ case T_BY:
+ case T_TO:
+ case T_WITH:
+ case T_EXP:
+ case T_MACRO_PUNCT:
+ return TC_BINOP;
+
+ case T_COMMA:
+ return TC_COMMA;
+ }
+
+ NOT_REACHED ();
+}
+
+/* Appends a syntax representation of the tokens in MTS to S. If OFS and LEN
+ are nonnull, sets OFS[i] to the offset within S of the start of token 'i' in
+ MTS and LEN[i] to its length. OFS[i] + LEN[i] is not necessarily OFS[i + 1]
+ because some tokens are separated by white space. */
+void
+macro_tokens_to_representation (struct macro_tokens *mts, struct string *s,
+ size_t *ofs, size_t *len)
+{
+ assert ((ofs != NULL) == (len != NULL));
+
+ if (!mts->n)
+ return;
+
+ for (size_t i = 0; i < mts->n; i++)
+ {
+ if (i > 0)
+ {
+ enum token_type prev = mts->mts[i - 1].token.type;
+ enum token_type next = mts->mts[i].token.type;
+
+ if (prev == T_ENDCMD)
+ ds_put_byte (s, '\n');
+ else
+ {
+ enum token_class pc = classify_token (prev);
+ enum token_class nc = classify_token (next);
+ if (needs_space (pc, nc))
+ ds_put_byte (s, ' ');
+ }
+ }
+
+ if (ofs)
+ ofs[i] = s->ss.length;
+ macro_token_to_representation (&mts->mts[i], s);
+ if (len)
+ len[i] = s->ss.length - ofs[i];
+ }
+}
+
+void
+macro_destroy (struct macro *m)
+{
+ if (!m)
+ return;
+
+ free (m->name);
+ free (m->file_name);
+ for (size_t i = 0; i < m->n_params; i++)
+ {
+ struct macro_param *p = &m->params[i];
+ free (p->name);
+
+ macro_tokens_uninit (&p->def);
+
+ switch (p->arg_type)
+ {
+ case ARG_N_TOKENS:
+ break;
+
+ case ARG_CHAREND:
+ token_uninit (&p->charend);
+ break;
+
+ case ARG_ENCLOSE:
+ token_uninit (&p->enclose[0]);
+ token_uninit (&p->enclose[1]);
+ break;
+
+ case ARG_CMDEND:
+ break;
+ }
+ }
+ free (m->params);
+ macro_tokens_uninit (&m->body);
+ free (m);
+}
+\f
+struct macro_set *
+macro_set_create (void)
+{
+ struct macro_set *set = xmalloc (sizeof *set);
+ *set = (struct macro_set) {
+ .macros = HMAP_INITIALIZER (set->macros),
+ };
+ return set;
+}
+
+void
+macro_set_destroy (struct macro_set *set)
+{
+ if (!set)
+ return;
+
+ struct macro *macro, *next;
+ HMAP_FOR_EACH_SAFE (macro, next, struct macro, hmap_node, &set->macros)
+ {
+ hmap_delete (&set->macros, ¯o->hmap_node);
+ macro_destroy (macro);
+ }
+ hmap_destroy (&set->macros);
+ free (set);
+}
+
+static unsigned int
+hash_macro_name (const char *name)
+{
+ return utf8_hash_case_string (name, 0);
+}
+
+static struct macro *
+macro_set_find__ (struct macro_set *set, const char *name)
+{
+ if (macro_set_is_empty (set))
+ return NULL;
+
+ struct macro *macro;
+ HMAP_FOR_EACH_WITH_HASH (macro, struct macro, hmap_node,
+ hash_macro_name (name), &set->macros)
+ if (!utf8_strcasecmp (macro->name, name))
+ return macro;
+
+ return NULL;
+}
+
+const struct macro *
+macro_set_find (const struct macro_set *set, const char *name)
+{
+ return macro_set_find__ (CONST_CAST (struct macro_set *, set), name);
+}
+
+/* Adds M to SET. M replaces any existing macro with the same name. Takes
+ ownership of M. */
+void
+macro_set_add (struct macro_set *set, struct macro *m)
+{
+ struct macro *victim = macro_set_find__ (set, m->name);
+ if (victim)
+ {
+ hmap_delete (&set->macros, &victim->hmap_node);
+ macro_destroy (victim);
+ }
+
+ hmap_insert (&set->macros, &m->hmap_node, hash_macro_name (m->name));
+}
+\f
+/* Macro expander. */
+
+enum me_state
+ {
+ /* Error state. */
+ ME_ERROR,
+
+ /* Accumulating tokens in me->params toward the end of any type of
+ argument. */
+ ME_ARG,
+
+ /* Expecting the opening delimiter of an ARG_ENCLOSE argument. */
+ ME_ENCLOSE,
+
+ /* Expecting a keyword for a keyword argument. */
+ ME_KEYWORD,
+
+ /* Expecting an equal sign for a keyword argument. */
+ ME_EQUALS,
+
+ /* Macro fully parsed and ready for expansion. */
+ ME_FINISHED,
+ };
+
+/* Macro expander.
+
+ Macro expansion has two phases. The first phase uses an FSM driven by
+ macro_expander_create() and macro_expander_add() to identify the macro being
+ called and obtain its arguments. 'state' identifies the FSM state.
+
+ The second phase is macro expansion via macro_expander_get_expansion(). */
+struct macro_expander
+ {
+ const struct macro_set *macros;
+ const struct macro *macro;
+
+ enum me_state state;
+ size_t n_tokens;
+
+ struct macro_tokens **args;
+ const struct macro_param *param; /* Parameter currently being parsed. */
+ };
+
+/* Completes macro expansion by initializing arguments that weren't supplied to
+ their defaults. */
+static int
+me_finished (struct macro_expander *me)
+{
+ me->state = ME_FINISHED;
+ for (size_t i = 0; i < me->macro->n_params; i++)
+ if (!me->args[i])
+ me->args[i] = &me->macro->params[i].def;
+ return me->n_tokens;
+}
+
+static int
+me_next_arg (struct macro_expander *me)
+{
+ if (!me->param)
+ {
+ assert (!me->macro->n_params);
+ return me_finished (me);
+ }
+ else if (me->param->positional)
+ {
+ me->param++;
+ if (me->param >= &me->macro->params[me->macro->n_params])
+ return me_finished (me);
+ else
+ {
+ me->state = (!me->param->positional ? ME_KEYWORD
+ : me->param->arg_type == ARG_ENCLOSE ? ME_ENCLOSE
+ : ME_ARG);
+ return 0;
+ }
+ }
+ else
+ {
+ for (size_t i = 0; i < me->macro->n_params; i++)
+ if (!me->args[i])
+ {
+ me->state = ME_KEYWORD;
+ return 0;
+ }
+ return me_finished (me);
+ }
+}
+
+static int
+me_error (struct macro_expander *me)
+{
+ me->state = ME_ERROR;
+ return -1;
+}
+
+static int
+me_add_arg (struct macro_expander *me, const struct macro_token *mt)
+{
+ const struct macro_param *p = me->param;
+
+ const struct token *token = &mt->token;
+ if ((token->type == T_ENDCMD || token->type == T_STOP)
+ && p->arg_type != ARG_CMDEND)
+ {
+ msg (SE, _("Unexpected end of command reading argument %s "
+ "to macro %s."), me->param->name, me->macro->name);
+
+ return me_error (me);
+ }
+
+ me->n_tokens++;
+
+ struct macro_tokens **argp = &me->args[p - me->macro->params];
+ if (!*argp)
+ *argp = xzalloc (sizeof **argp);
+ struct macro_tokens *arg = *argp;
+ if (p->arg_type == ARG_N_TOKENS)
+ {
+ macro_tokens_add (arg, mt);
+ if (arg->n >= p->n_tokens)
+ return me_next_arg (me);
+ return 0;
+ }
+ else if (p->arg_type == ARG_CMDEND)
+ {
+ if (token->type == T_ENDCMD || token->type == T_STOP)
+ return me_next_arg (me);
+ macro_tokens_add (arg, mt);
+ return 0;
+ }
+ else
+ {
+ const struct token *end
+ = p->arg_type == ARG_CHAREND ? &p->charend : &p->enclose[1];
+ if (token_equal (token, end))
+ return me_next_arg (me);
+ macro_tokens_add (arg, mt);
+ return 0;
+ }
+}
+
+static int
+me_expected (struct macro_expander *me, const struct macro_token *actual,
+ const struct token *expected)
+{
+ const struct substring actual_s
+ = (actual->representation.length ? actual->representation
+ : ss_cstr (_("<end of input>")));
+ char *expected_s = token_to_string (expected);
+ msg (SE, _("Found `%.*s' while expecting `%s' reading argument %s "
+ "to macro %s."),
+ (int) actual_s.length, actual_s.string, expected_s,
+ me->param->name, me->macro->name);
+ free (expected_s);
+
+ return me_error (me);
+}
+
+static int
+me_enclose (struct macro_expander *me, const struct macro_token *mt)
+{
+ const struct token *token = &mt->token;
+ me->n_tokens++;
+
+ if (token_equal (&me->param->enclose[0], token))
+ {
+ me->state = ME_ARG;
+ return 0;
+ }
+
+ return me_expected (me, mt, &me->param->enclose[0]);
+}
+
+static const struct macro_param *
+macro_find_parameter_by_name (const struct macro *m, struct substring name)
+{
+ ss_ltrim (&name, ss_cstr ("!"));
+
+ for (size_t i = 0; i < m->n_params; i++)
+ {
+ const struct macro_param *p = &m->params[i];
+ struct substring p_name = ss_cstr (p->name + 1);
+ if (!utf8_strncasecmp (p_name.string, p_name.length,
+ name.string, name.length))
+ return p;
+ }
+ return NULL;
+}
+
+static int
+me_keyword (struct macro_expander *me, const struct macro_token *mt)
+{
+ const struct token *token = &mt->token;
+ if (token->type != T_ID)
+ return me_finished (me);
+
+ const struct macro_param *p = macro_find_parameter_by_name (me->macro,
+ token->string);
+ if (p)
+ {
+ size_t arg_index = p - me->macro->params;
+ me->param = p;
+ if (me->args[arg_index])
+ {
+ msg (SE,
+ _("Argument %s multiply specified in call to macro %s."),
+ p->name, me->macro->name);
+ return me_error (me);
+ }
+
+ me->n_tokens++;
+ me->state = ME_EQUALS;
+ return 0;
+ }
+
+ return me_finished (me);
+}
+
+static int
+me_equals (struct macro_expander *me, const struct macro_token *mt)
+{
+ const struct token *token = &mt->token;
+ me->n_tokens++;
+
+ if (token->type == T_EQUALS)
+ {
+ me->state = ME_ARG;
+ return 0;
+ }
+
+ return me_expected (me, mt, &(struct token) { .type = T_EQUALS });
+}
+
+/* If TOKEN is the first token of a call to a macro in MACROS, create a new
+ macro expander, initializes *MEP to it. Returns 0 if more tokens are needed
+ and should be added via macro_expander_add() or 1 if the caller should next
+ call macro_expander_get_expansion().
+
+ If TOKEN is not the first token of a macro call, returns -1 and sets *MEP to
+ NULL. */
+int
+macro_expander_create (const struct macro_set *macros,
+ const struct token *token,
+ struct macro_expander **mep)
+{
+ const struct macro *macro = (token->type == T_ID || token->type == T_MACRO_ID
+ ? macro_set_find (macros, token->string.string)
+ : NULL);
+ if (!macro)
+ {
+ *mep = NULL;
+ return -1;
+ }
+
+ struct macro_expander *me = xmalloc (sizeof *me);
+ *me = (struct macro_expander) {
+ .macros = macros,
+ .macro = macro,
+ .n_tokens = 1,
+ .state = (!macro->n_params ? ME_FINISHED
+ : !macro->params[0].positional ? ME_KEYWORD
+ : macro->params[0].arg_type == ARG_ENCLOSE ? ME_ENCLOSE
+ : ME_ARG),
+ .args = macro->n_params ? xcalloc (macro->n_params, sizeof *me->args) : NULL,
+ .param = macro->params,
+ };
+ *mep = me;
+
+ return me->state == ME_FINISHED ? 1 : 0;
+}
+
+void
+macro_expander_destroy (struct macro_expander *me)
+{
+ if (!me)
+ return;
+
+ for (size_t i = 0; i < me->macro->n_params; i++)
+ {
+ struct macro_tokens *a = me->args[i];
+ if (a && a != &me->macro->params[i].def)
+ {
+ macro_tokens_uninit (a);
+ free (a);
+ }
+ }
+ free (me->args);
+ free (me);
+}
+
+/* Adds TOKEN to the collection of tokens in ME that potentially need to be
+ macro expanded.
+
+ Returns -1 if the tokens added do not actually invoke a macro. The caller
+ should consume the first token without expanding it. (Later tokens might
+ invoke a macro so it's best to feed the second token into a new expander.)
+
+ Returns 0 if the macro expander needs more tokens, for macro arguments or to
+ decide whether this is actually a macro invocation. The caller should call
+ macro_expander_add() again with the next token.
+
+ Returns a positive number to indicate that the returned number of tokens
+ invoke a macro. The number returned might be less than the number of tokens
+ added because it can take a few tokens of lookahead to determine whether the
+ macro invocation is finished. The caller should call
+ macro_expander_get_expansion() to obtain the expansion. */
+int
+macro_expander_add (struct macro_expander *me, const struct macro_token *mt)
+{
+ switch (me->state)
+ {
+ case ME_ERROR:
+ return -1;
+
+ case ME_ARG:
+ return me_add_arg (me, mt);
+
+ case ME_ENCLOSE:
+ return me_enclose (me, mt);
+
+ case ME_KEYWORD:
+ return me_keyword (me, mt);
+
+ case ME_EQUALS:
+ return me_equals (me, mt);
+
+ default:
+ NOT_REACHED ();
+ }
+}
+
+/* Each argument to a macro function is one of:
+
+ - A quoted string or other single literal token.
+
+ - An argument to the macro being expanded, e.g. !1 or a named argument.
+
+ - !*.
+
+ - A function invocation.
+
+ Each function invocation yields a character sequence to be turned into a
+ sequence of tokens. The case where that character sequence is a single
+ quoted string is an important special case.
+*/
+struct parse_macro_function_ctx
+ {
+ const struct macro_token *input;
+ size_t n_input;
+ int nesting_countdown;
+ enum segmenter_mode segmenter_mode;
+ const struct macro_set *macros;
+ const struct macro_expander *me;
+ const struct macro_expansion_stack *stack;
+ struct string_map *vars;
+ bool *expand;
+ };
+
+static void
+macro_expand (const struct macro_tokens *, int nesting_countdown,
+ enum segmenter_mode segmenter_mode, const struct macro_set *,
+ const struct macro_expander *, struct string_map *vars,
+ const struct macro_expansion_stack *stack,
+ bool *expand, bool *break_,
+ struct macro_tokens *exp);
+
+static bool
+expand_macro_function (struct parse_macro_function_ctx *ctx,
+ struct string *output, size_t *input_consumed);
+
+/* Returns true if the pair of tokens starting at offset OFS within MTS are !*,
+ false otherwise. */
+static bool
+is_bang_star (const struct macro_token *mts, size_t n, size_t ofs)
+{
+ return (ofs + 1 < n
+ && mts[ofs].token.type == T_MACRO_ID
+ && ss_equals (mts[ofs].token.string, ss_cstr ("!"))
+ && mts[ofs + 1].token.type == T_ASTERISK);
+}
+
+static size_t
+parse_function_arg (struct parse_macro_function_ctx *ctx,
+ size_t i, struct string *farg)
+{
+ const struct macro_token *tokens = ctx->input;
+ const struct token *token = &tokens[i].token;
+ if (token->type == T_MACRO_ID)
+ {
+ const struct macro_param *param = macro_find_parameter_by_name (
+ ctx->me->macro, token->string);
+ if (param)
+ {
+ size_t param_idx = param - ctx->me->macro->params;
+ const struct macro_tokens *marg = ctx->me->args[param_idx];
+ for (size_t i = 0; i < marg->n; i++)
+ {
+ if (i)
+ ds_put_byte (farg, ' ');
+ ds_put_substring (farg, marg->mts[i].representation);
+ }
+ return 1;
+ }
+
+ if (is_bang_star (ctx->input, ctx->n_input, i))
+ {
+ for (size_t i = 0; i < ctx->me->macro->n_params; i++)
+ {
+ if (!ctx->me->macro->params[i].positional)
+ break;
+
+ const struct macro_tokens *marg = ctx->me->args[i];
+ for (size_t j = 0; j < marg->n; j++)
+ {
+ if (i || j)
+ ds_put_byte (farg, ' ');
+ ds_put_substring (farg, marg->mts[j].representation);
+ }
+ }
+ return 2;
+ }
+
+ if (ctx->vars)
+ {
+ const char *value = string_map_find__ (ctx->vars,
+ token->string.string,
+ token->string.length);
+ if (value)
+ {
+ ds_put_cstr (farg, value);
+ return 1;
+ }
+ }
+
+ struct parse_macro_function_ctx subctx = {
+ .input = &ctx->input[i],
+ .n_input = ctx->n_input - i,
+ .nesting_countdown = ctx->nesting_countdown,
+ .segmenter_mode = ctx->segmenter_mode,
+ .macros = ctx->macros,
+ .me = ctx->me,
+ .stack = ctx->stack,
+ .vars = ctx->vars,
+ .expand = ctx->expand,
+ };
+ size_t subinput_consumed;
+ if (expand_macro_function (&subctx, farg, &subinput_consumed))
+ return subinput_consumed;
+ }
+
+ ds_put_substring (farg, tokens[i].representation);
+ return 1;
+}
+
+static bool
+parse_macro_function (struct parse_macro_function_ctx *ctx,
+ struct string_array *args,
+ struct substring function,
+ int min_args, int max_args,
+ size_t *input_consumed)
+{
+ const struct macro_token *tokens = ctx->input;
+ size_t n_tokens = ctx->n_input;
+
+ if (!n_tokens
+ || tokens[0].token.type != T_MACRO_ID
+ || !ss_equals_case (tokens[0].token.string, function)) /* XXX abbrevs allowed */
+ return false;
+
+ if (n_tokens < 2 || tokens[1].token.type != T_LPAREN)
+ {
+ macro_error (ctx->stack, n_tokens > 1 ? &tokens[1] : NULL,
+ _("`(' expected following %s."), function.string);
+ return false;
+ }
+
+ string_array_init (args);
+
+ for (size_t i = 2;; )
+ {
+ if (i >= n_tokens)
+ goto unexpected_end;
+ if (tokens[i].token.type == T_RPAREN)
+ {
+ *input_consumed = i + 1;
+ if (args->n < min_args || args->n > max_args)
+ {
+ macro_error (ctx->stack, &tokens[i],
+ _("Wrong number of arguments to macro function %s."),
+ function.string);
+ goto error;
+ }
+ return true;
+ }
+
+ struct string s = DS_EMPTY_INITIALIZER;
+ i += parse_function_arg (ctx, i, &s);
+ if (i >= n_tokens)
+ {
+ ds_destroy (&s);
+ goto unexpected_end;
+ }
+ string_array_append_nocopy (args, ds_steal_cstr (&s));
+
+ if (tokens[i].token.type == T_COMMA)
+ i++;
+ else if (tokens[i].token.type != T_RPAREN)
+ {
+ macro_error (ctx->stack, &tokens[i],
+ _("`,' or `)' expected in call to macro function %s."),
+ function.string);
+ goto error;
+ }
+ }
+
+unexpected_end:
+ macro_error (ctx->stack, NULL, _("Missing `)' in call to macro function %s."),
+ function.string);
+ /* Fall through. */
+error:
+ string_array_destroy (args);
+ return false;
+}
+
+static bool
+unquote_string (const char *s, enum segmenter_mode segmenter_mode,
+ struct string *content)
+{
+ struct string_lexer slex;
+ string_lexer_init (&slex, s, strlen (s), segmenter_mode, true);
+
+ struct token token1;
+ if (!string_lexer_next (&slex, &token1))
+ return false;
+
+ if (token1.type != T_STRING)
+ {
+ token_uninit (&token1);
+ return false;
+ }
+
+ struct token token2;
+ if (string_lexer_next (&slex, &token2))
+ {
+ token_uninit (&token1);
+ token_uninit (&token2);
+ return false;
+ }
+
+ ds_put_substring (content, token1.string);
+ token_uninit (&token1);
+ return true;
+}
+
+static const char *
+unquote_string_in_place (const char *s, enum segmenter_mode segmenter_mode,
+ struct string *tmp)
+{
+ ds_init_empty (tmp);
+ return unquote_string (s, segmenter_mode, tmp) ? ds_cstr (tmp) : s;
+}
+
+static bool
+parse_integer (const char *s, int *np)
+{
+ errno = 0;
+
+ char *tail;
+ long int n = strtol (s, &tail, 10);
+ *np = n < INT_MIN ? INT_MIN : n > INT_MAX ? INT_MAX : n;
+ tail += strspn (tail, CC_SPACES);
+ return *tail == '\0' && errno != ERANGE && n == *np;
+}
+
+static bool
+expand_macro_function (struct parse_macro_function_ctx *ctx,
+ struct string *output,
+ size_t *input_consumed)
+{
+ struct string_array args;
+
+ if (parse_macro_function (ctx, &args, ss_cstr ("!LENGTH"), 1, 1,
+ input_consumed))
+ ds_put_format (output, "%zu", strlen (args.strings[0]));
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!BLANKS"), 1, 1,
+ input_consumed))
+ {
+ int n;
+ if (!parse_integer (args.strings[0], &n))
+ {
+ macro_error (ctx->stack, NULL,
+ _("Argument to !BLANKS must be non-negative integer "
+ "(not \"%s\")."), args.strings[0]);
+ string_array_destroy (&args);
+ return false;
+ }
+
+ ds_put_byte_multiple (output, ' ', n);
+ }
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!CONCAT"), 1, INT_MAX,
+ input_consumed))
+ {
+ for (size_t i = 0; i < args.n; i++)
+ if (!unquote_string (args.strings[i], ctx->segmenter_mode, output))
+ ds_put_cstr (output, args.strings[i]);
+ }
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!HEAD"), 1, 1,
+ input_consumed))
+ {
+ struct string tmp;
+ const char *s = unquote_string_in_place (args.strings[0],
+ ctx->segmenter_mode, &tmp);
+
+ struct macro_tokens mts = { .n = 0 };
+ macro_tokens_from_string__ (&mts, ss_cstr (s), ctx->segmenter_mode,
+ ctx->stack);
+ if (mts.n > 0)
+ ds_put_substring (output, mts.mts[0].representation);
+ macro_tokens_uninit (&mts);
+ ds_destroy (&tmp);
+ }
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!INDEX"), 2, 2,
+ input_consumed))
+ {
+ const char *haystack = args.strings[0];
+ const char *needle = strstr (haystack, args.strings[1]);
+ ds_put_format (output, "%zu", needle ? needle - haystack + 1 : 0);
+ }
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!QUOTE"), 1, 1,
+ input_consumed))
+ {
+ if (unquote_string (args.strings[0], ctx->segmenter_mode, NULL))
+ ds_put_cstr (output, args.strings[0]);
+ else
+ {
+ ds_extend (output, strlen (args.strings[0]) + 2);
+ ds_put_byte (output, '\'');
+ for (const char *p = args.strings[0]; *p; p++)
+ {
+ if (*p == '\'')
+ ds_put_byte (output, '\'');
+ ds_put_byte (output, *p);
+ }
+ ds_put_byte (output, '\'');
+ }
+ }
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!SUBSTR"), 2, 3,
+ input_consumed))
+ {
+ int start;
+ if (!parse_integer (args.strings[1], &start) || start < 1)
+ {
+ macro_error (ctx->stack, NULL,
+ _("Second argument of !SUBSTR must be "
+ "positive integer (not \"%s\")."),
+ args.strings[1]);
+ string_array_destroy (&args);
+ return false;
+ }
+
+ int count = INT_MAX;
+ if (args.n > 2 && (!parse_integer (args.strings[2], &count) || count < 0))
+ {
+ macro_error (ctx->stack, NULL,
+ _("Third argument of !SUBSTR must be "
+ "non-negative integer (not \"%s\")."),
+ args.strings[2]);
+ string_array_destroy (&args);
+ return false;
+ }
+
+ struct substring s = ss_cstr (args.strings[0]);
+ ds_put_substring (output, ss_substr (s, start - 1, count));
+ }
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!TAIL"), 1, 1,
+ input_consumed))
+ {
+ struct string tmp;
+ const char *s = unquote_string_in_place (args.strings[0],
+ ctx->segmenter_mode, &tmp);
+
+ struct macro_tokens mts = { .n = 0 };
+ macro_tokens_from_string__ (&mts, ss_cstr (s), ctx->segmenter_mode,
+ ctx->stack);
+ if (mts.n > 1)
+ {
+ struct macro_tokens tail = { .mts = mts.mts + 1, .n = mts.n - 1 };
+ macro_tokens_to_representation (&tail, output, NULL, NULL);
+ }
+ macro_tokens_uninit (&mts);
+ ds_destroy (&tmp);
+ }
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!UNQUOTE"), 1, 1,
+ input_consumed))
+ {
+ if (!unquote_string (args.strings[0], ctx->segmenter_mode, output))
+ ds_put_cstr (output, args.strings[0]);
+ }
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!UPCASE"), 1, 1,
+ input_consumed))
+ {
+ struct string tmp;
+ const char *s = unquote_string_in_place (args.strings[0],
+ ctx->segmenter_mode, &tmp);
+ char *upper = utf8_to_upper (s);
+ ds_put_cstr (output, upper);
+ free (upper);
+ ds_destroy (&tmp);
+ }
+ else if (parse_macro_function (ctx, &args, ss_cstr ("!EVAL"), 1, 1,
+ input_consumed))
+ {
+ struct macro_tokens mts = { .n = 0 };
+ macro_tokens_from_string__ (&mts, ss_cstr (args.strings[0]),
+ ctx->segmenter_mode, ctx->stack);
+ struct macro_tokens exp = { .n = 0 };
+ macro_expand (&mts, ctx->nesting_countdown - 1, ctx->segmenter_mode,
+ ctx->macros, ctx->me, ctx->vars,
+ &(struct macro_expansion_stack) {
+ .name = "!EVAL",
+ .next = ctx->stack,
+ }, ctx->expand, NULL, &exp);
+ macro_tokens_to_representation (&exp, output, NULL, NULL);
+ macro_tokens_uninit (&exp);
+ macro_tokens_uninit (&mts);
+ }
+ else if (ctx->n_input > 0
+ && ctx->input[0].token.type == T_MACRO_ID
+ && ss_equals_case (ctx->input[0].token.string, ss_cstr ("!NULL")))
+ {
+ *input_consumed = 1;
+ return true;
+ }
+ else
+ return false;
+
+ string_array_destroy (&args);
+ return true;
+}
+
+struct expr_context
+ {
+ int nesting_countdown;
+ enum segmenter_mode segmenter_mode;
+ const struct macro_set *macros;
+ const struct macro_expander *me;
+ const struct macro_expansion_stack *stack;
+ struct string_map *vars;
+ bool *expand;
+ };
+
+static char *macro_evaluate_or (const struct expr_context *ctx,
+ const struct macro_token **tokens,
+ const struct macro_token *end);
+
+static char *
+macro_evaluate_literal (const struct expr_context *ctx,
+ const struct macro_token **tokens,
+ const struct macro_token *end)
+{
+ const struct macro_token *p = *tokens;
+ if (p >= end)
+ return NULL;
+ if (p->token.type == T_LPAREN)
+ {
+ p++;
+ char *value = macro_evaluate_or (ctx, &p, end);
+ if (!value)
+ return NULL;
+ if (p >= end || p->token.type != T_RPAREN)
+ {
+ free (value);
+ macro_error (ctx->stack, p < end ? p : NULL,
+ _("Expecting ')' in macro expression."));
+ return NULL;
+ }
+ p++;
+ *tokens = p;
+ return value;
+ }
+ else if (p->token.type == T_RPAREN)
+ {
+ macro_error (ctx->stack, p, _("Expecting literal or function invocation "
+ "in macro expression."));
+ return NULL;
+ }
+
+ struct parse_macro_function_ctx fctx = {
+ .input = p,
+ .n_input = end - p,
+ .nesting_countdown = ctx->nesting_countdown,
+ .segmenter_mode = ctx->segmenter_mode,
+ .macros = ctx->macros,
+ .me = ctx->me,
+ .stack = ctx->stack,
+ .vars = ctx->vars,
+ .expand = ctx->expand,
+ };
+ struct string function_output = DS_EMPTY_INITIALIZER;
+ size_t function_consumed = parse_function_arg (&fctx, 0, &function_output);
+ struct string unquoted = DS_EMPTY_INITIALIZER;
+ if (unquote_string (ds_cstr (&function_output), ctx->segmenter_mode,
+ &unquoted))
+ {
+ ds_swap (&function_output, &unquoted);
+ ds_destroy (&unquoted);
+ }
+ *tokens = p + function_consumed;
+ return ds_steal_cstr (&function_output);
+}
+
+/* Returns true if MT is valid as a macro operator. Only operators written as
+ symbols (e.g. <>) are usable in macro expressions, not operator written as
+ letters (e.g. EQ). */
+static bool
+is_macro_operator (const struct macro_token *mt)
+{
+ return (mt->representation.length > 0
+ && !c_isalpha (mt->representation.string[0]));
+}
+
+static enum token_type
+parse_relational_op (const struct macro_token *mt)
+{
+ switch (mt->token.type)
+ {
+ case T_EQUALS:
+ return T_EQ;
+
+ case T_NE:
+ case T_LT:
+ case T_GT:
+ case T_LE:
+ case T_GE:
+ return is_macro_operator (mt) ? mt->token.type : T_STOP;
+
+ case T_MACRO_ID:
+ return (ss_equals_case (mt->token.string, ss_cstr ("!EQ")) ? T_EQ
+ : ss_equals_case (mt->token.string, ss_cstr ("!NE")) ? T_NE
+ : ss_equals_case (mt->token.string, ss_cstr ("!LT")) ? T_LT
+ : ss_equals_case (mt->token.string, ss_cstr ("!GT")) ? T_GT
+ : ss_equals_case (mt->token.string, ss_cstr ("!LE")) ? T_LE
+ : ss_equals_case (mt->token.string, ss_cstr ("!GE")) ? T_GE
+ : T_STOP);
+
+ default:
+ return T_STOP;
+ }
+}
+
+static char *
+macro_evaluate_relational (const struct expr_context *ctx,
+ const struct macro_token **tokens,
+ const struct macro_token *end)
+{
+ const struct macro_token *p = *tokens;
+ char *lhs = macro_evaluate_literal (ctx, &p, end);
+ if (!lhs)
+ return NULL;
+
+ enum token_type op = p >= end ? T_STOP : parse_relational_op (p);
+ if (op == T_STOP)
+ {
+ *tokens = p;
+ return lhs;
+ }
+ p++;
+
+ char *rhs = macro_evaluate_literal (ctx, &p, end);
+ if (!rhs)
+ {
+ free (lhs);
+ return NULL;
+ }
+
+ struct string lhs_tmp, rhs_tmp;
+ int cmp = strcmp (unquote_string_in_place (lhs, ctx->segmenter_mode,
+ &lhs_tmp),
+ unquote_string_in_place (rhs, ctx->segmenter_mode,
+ &rhs_tmp));
+ ds_destroy (&lhs_tmp);
+ ds_destroy (&rhs_tmp);
+
+ free (lhs);
+ free (rhs);
+
+ bool b = (op == T_EQUALS || op == T_EQ ? !cmp
+ : op == T_NE ? cmp
+ : op == T_LT ? cmp < 0
+ : op == T_GT ? cmp > 0
+ : op == T_LE ? cmp <= 0
+ : /* T_GE */ cmp >= 0);
+
+ *tokens = p;
+ return xstrdup (b ? "1" : "0");
+}
+
+static char *
+macro_evaluate_not (const struct expr_context *ctx,
+ const struct macro_token **tokens,
+ const struct macro_token *end)
+{
+ const struct macro_token *p = *tokens;
+
+ unsigned int negations = 0;
+ while (p < end
+ && (ss_equals_case (p->representation, ss_cstr ("!NOT"))
+ || ss_equals (p->representation, ss_cstr ("~"))))
+ {
+ p++;
+ negations++;
+ }
+
+ char *operand = macro_evaluate_relational (ctx, &p, end);
+ if (!operand || !negations)
+ {
+ *tokens = p;
+ return operand;
+ }
+
+ bool b = strcmp (operand, "0") ^ (negations & 1);
+ free (operand);
+ *tokens = p;
+ return xstrdup (b ? "1" : "0");
+}
+
+static char *
+macro_evaluate_and (const struct expr_context *ctx,
+ const struct macro_token **tokens,
+ const struct macro_token *end)
+{
+ const struct macro_token *p = *tokens;
+ char *lhs = macro_evaluate_not (ctx, &p, end);
+ if (!lhs)
+ return NULL;
+
+ while (p < end
+ && (ss_equals_case (p->representation, ss_cstr ("!AND"))
+ || ss_equals (p->representation, ss_cstr ("&"))))
+ {
+ p++;
+ char *rhs = macro_evaluate_not (ctx, &p, end);
+ if (!rhs)
+ {
+ free (lhs);
+ return NULL;
+ }
+
+ bool b = strcmp (lhs, "0") && strcmp (rhs, "0");
+ free (lhs);
+ free (rhs);
+ lhs = xstrdup (b ? "1" : "0");
+ }
+ *tokens = p;
+ return lhs;
+}
+
+static char *
+macro_evaluate_or (const struct expr_context *ctx,
+ const struct macro_token **tokens,
+ const struct macro_token *end)
+{
+ const struct macro_token *p = *tokens;
+ char *lhs = macro_evaluate_and (ctx, &p, end);
+ if (!lhs)
+ return NULL;
+
+ while (p < end
+ && (ss_equals_case (p->representation, ss_cstr ("!OR"))
+ || ss_equals (p->representation, ss_cstr ("|"))))
+ {
+ p++;
+ char *rhs = macro_evaluate_and (ctx, &p, end);
+ if (!rhs)
+ {
+ free (lhs);
+ return NULL;
+ }
+
+ bool b = strcmp (lhs, "0") || strcmp (rhs, "0");
+ free (lhs);
+ free (rhs);
+ lhs = xstrdup (b ? "1" : "0");
+ }
+ *tokens = p;
+ return lhs;
+}
+
+static char *
+macro_evaluate_expression (const struct macro_token **tokens, size_t n_tokens,
+ int nesting_countdown,
+ enum segmenter_mode segmenter_mode,
+ const struct macro_set *macros,
+ const struct macro_expander *me,
+ const struct macro_expansion_stack *stack,
+ struct string_map *vars, bool *expand)
+{
+ const struct expr_context ctx = {
+ .nesting_countdown = nesting_countdown,
+ .segmenter_mode = segmenter_mode,
+ .macros = macros,
+ .me = me,
+ .stack = stack,
+ .vars = vars,
+ .expand = expand,
+ };
+ return macro_evaluate_or (&ctx, tokens, *tokens + n_tokens);
+}
+
+static bool
+macro_evaluate_number (const struct macro_token **tokens, size_t n_tokens,
+ int nesting_countdown,
+ enum segmenter_mode segmenter_mode,
+ const struct macro_set *macros,
+ const struct macro_expander *me,
+ const struct macro_expansion_stack *stack,
+ struct string_map *vars,
+ bool *expand, double *number)
+{
+ char *s = macro_evaluate_expression (tokens, n_tokens, nesting_countdown,
+ segmenter_mode, macros, me, stack, vars,
+ expand);
+ if (!s)
+ return false;
+
+ struct macro_tokens mts = { .n = 0 };
+ macro_tokens_from_string__ (&mts, ss_cstr (s), segmenter_mode, stack);
+ if (mts.n != 1 || !token_is_number (&mts.mts[0].token))
+ {
+ macro_error (stack, mts.n > 0 ? &mts.mts[0] : NULL,
+ _("Macro expression must evaluate to "
+ "a number (not \"%s\")."), s);
+ free (s);
+ macro_tokens_uninit (&mts);
+ return false;
+ }
+
+ *number = token_number (&mts.mts[0].token);
+ free (s);
+ macro_tokens_uninit (&mts);
+ return true;
+}
+
+static const struct macro_token *
+find_ifend_clause (const struct macro_token *p, const struct macro_token *end)
+{
+ size_t nesting = 0;
+ for (; p < end; p++)
+ {
+ if (p->token.type != T_MACRO_ID)
+ continue;
+
+ if (ss_equals_case (p->token.string, ss_cstr ("!IF")))
+ nesting++;
+ else if (ss_equals_case (p->token.string, ss_cstr ("!IFEND")))
+ {
+ if (!nesting)
+ return p;
+ nesting--;
+ }
+ else if (ss_equals_case (p->token.string, ss_cstr ("!ELSE")) && !nesting)
+ return p;
+ }
+ return NULL;
+}
+
+static size_t
+macro_expand_if (const struct macro_token *tokens, size_t n_tokens,
+ int nesting_countdown, enum segmenter_mode segmenter_mode,
+ const struct macro_set *macros,
+ const struct macro_expander *me,
+ const struct macro_expansion_stack *stack,
+ struct string_map *vars,
+ bool *expand, bool *break_, struct macro_tokens *exp)
+{
+ const struct macro_token *p = tokens;
+ const struct macro_token *end = tokens + n_tokens;
+
+ if (p >= end || !ss_equals_case (p->token.string, ss_cstr ("!IF")))
+ return 0;
+
+ p++;
+ char *result = macro_evaluate_expression (&p, end - p,
+ nesting_countdown, segmenter_mode,
+ macros, me,
+ stack, vars, expand);
+ if (!result)
+ return 0;
+ bool b = strcmp (result, "0");
+ free (result);
+
+ if (p >= end
+ || p->token.type != T_MACRO_ID
+ || !ss_equals_case (p->token.string, ss_cstr ("!THEN")))
+ {
+ macro_error (stack, p < end ? p : NULL,
+ _("!THEN expected in macro !IF construct."));
+ return 0;
+ }
+
+ const struct macro_token *start_then = p + 1;
+ const struct macro_token *end_then = find_ifend_clause (start_then, end);
+ if (!end_then)
+ {
+ macro_error (stack, NULL,
+ _("!ELSE or !IFEND expected in macro !IF construct."));
+ return 0;
+ }
+
+ const struct macro_token *start_else, *end_if;
+ if (ss_equals_case (end_then->token.string, ss_cstr ("!ELSE")))
+ {
+ start_else = end_then + 1;
+ end_if = find_ifend_clause (start_else, end);
+ if (!end_if
+ || !ss_equals_case (end_if->token.string, ss_cstr ("!IFEND")))
+ {
+ macro_error (stack, end_if ? end_if : NULL,
+ _("!IFEND expected in macro !IF construct."));
+ return 0;
+ }
+ }
+ else
+ {
+ start_else = NULL;
+ end_if = end_then;
+ }
+
+ const struct macro_token *start;
+ size_t n;
+ if (b)
+ {
+ start = start_then;
+ n = end_then - start_then;
+ }
+ else if (start_else)
+ {
+ start = start_else;
+ n = end_if - start_else;
+ }
+ else
+ {
+ start = NULL;
+ n = 0;
+ }
+
+ if (n)
+ {
+ struct macro_tokens mts = {
+ .mts = CONST_CAST (struct macro_token *, start),
+ .n = n,
+ };
+ macro_expand (&mts, nesting_countdown, segmenter_mode, macros, me, vars,
+ &(struct macro_expansion_stack) {
+ .name = "!IF",
+ .next = stack,
+ },
+ expand, break_, exp);
+ }
+ return (end_if + 1) - tokens;
+}
+
+static size_t
+macro_parse_let (const struct macro_token *tokens, size_t n_tokens,
+ int nesting_countdown, enum segmenter_mode segmenter_mode,
+ const struct macro_set *macros,
+ const struct macro_expander *me,
+ const struct macro_expansion_stack *stack,
+ struct string_map *vars, bool *expand)
+{
+ const struct macro_token *p = tokens;
+ const struct macro_token *end = tokens + n_tokens;
+
+ if (p >= end || !ss_equals_case (p->token.string, ss_cstr ("!LET")))
+ return 0;
+ p++;
+
+ if (p >= end || p->token.type != T_MACRO_ID)
+ {
+ macro_error (stack, p < end ? p : NULL,
+ _("Expected macro variable name following !LET."));
+ return 0;
+ }
+ const struct substring var_name = p->token.string;
+ if (is_macro_keyword (var_name)
+ || macro_find_parameter_by_name (me->macro, var_name))
+ {
+ macro_error (stack, p < end ? p : NULL,
+ _("Cannot use argument name or macro keyword "
+ "\"%.*s\" as !LET variable."),
+ (int) var_name.length, var_name.string);
+ return 0;
+ }
+ p++;
+
+ if (p >= end || p->token.type != T_EQUALS)
+ {
+ macro_error (stack, p < end ? p : NULL,
+ _("Expected `=' following !LET."));
+ return 0;
+ }
+ p++;
+
+ char *value = macro_evaluate_expression (&p, end - p, nesting_countdown,
+ segmenter_mode, macros, me, stack,
+ vars, expand);
+ if (!value)
+ return 0;
+
+ string_map_replace_nocopy (vars, ss_xstrdup (var_name), value);
+ return p - tokens;
+}
+
+static const struct macro_token *
+find_doend (const struct macro_expansion_stack *stack,
+ const struct macro_token *p, const struct macro_token *end)
+{
+ size_t nesting = 0;
+ for (; p < end; p++)
+ {
+ if (p->token.type != T_MACRO_ID)
+ continue;
+
+ if (ss_equals_case (p->token.string, ss_cstr ("!DO")))
+ nesting++;
+ else if (ss_equals_case (p->token.string, ss_cstr ("!DOEND")))
+ {
+ if (!nesting)
+ return p;
+ nesting--;
+ }
+ }
+ macro_error (stack, NULL, _("Missing !DOEND."));
+ return NULL;
+}
+
+static size_t
+macro_expand_do (const struct macro_token *tokens, size_t n_tokens,
+ int nesting_countdown, enum segmenter_mode segmenter_mode,
+ const struct macro_set *macros,
+ const struct macro_expander *me,
+ const struct macro_expansion_stack *stack,
+ struct string_map *vars,
+ bool *expand, struct macro_tokens *exp)
+{
+ const struct macro_token *p = tokens;
+ const struct macro_token *end = tokens + n_tokens;
+
+ if (p >= end || !ss_equals_case (p->token.string, ss_cstr ("!DO")))
+ return 0;
+ p++;
+
+ if (p >= end || p->token.type != T_MACRO_ID)
+ {
+ macro_error (stack, p < end ? p : NULL,
+ _("Expected macro variable name following !DO."));
+ return 0;
+ }
+ const struct substring var_name = p->token.string;
+ if (is_macro_keyword (var_name)
+ || macro_find_parameter_by_name (me->macro, var_name))
+ {
+ macro_error (stack, p, _("Cannot use argument name or macro "
+ "keyword as !DO variable."));
+ return 0;
+ }
+ p++;
+
+ struct macro_expansion_stack next_stack = {
+ .name = "!DO", .next = stack,
+ };
+ int miterate = settings_get_miterate ();
+ if (p < end && p->token.type == T_MACRO_ID
+ && ss_equals_case (p->token.string, ss_cstr ("!IN")))
+ {
+ p++;
+ char *list = macro_evaluate_expression (&p, end - p, nesting_countdown,
+ segmenter_mode, macros, me,
+ &next_stack, vars, expand);
+ if (!list)
+ return 0;
+
+ struct macro_tokens items = { .n = 0 };
+ macro_tokens_from_string__ (&items, ss_cstr (list), segmenter_mode,
+ stack);
+ free (list);
+
+ const struct macro_token *do_end = find_doend (stack, p, end);
+ if (!do_end)
+ {
+ macro_tokens_uninit (&items);
+ return 0;
+ }
+
+ const struct macro_tokens inner = {
+ .mts = CONST_CAST (struct macro_token *, p),
+ .n = do_end - p
+ };
+ for (size_t i = 0; i < items.n; i++)
+ {
+ if (i >= miterate)
+ {
+ macro_error (stack, NULL,
+ _("!DO loop over list exceeded "
+ "maximum number of iterations %d. "
+ "(Use SET MITERATE to change the limit.)"),
+ miterate);
+ break;
+ }
+ string_map_replace_nocopy (vars, ss_xstrdup (var_name),
+ ss_xstrdup (items.mts[i].representation));
+
+ bool break_ = false;
+ macro_expand (&inner, nesting_countdown, segmenter_mode, macros,
+ me, vars, &next_stack, expand, &break_, exp);
+ if (break_)
+ break;
+ }
+ return do_end - tokens + 1;
+ }
+ else if (p < end && p->token.type == T_EQUALS)
+ {
+ p++;
+ double first;
+ if (!macro_evaluate_number (&p, end - p, nesting_countdown,
+ segmenter_mode, macros, me, &next_stack,
+ vars, expand, &first))
+ return 0;
+
+ if (p >= end || p->token.type != T_MACRO_ID
+ || !ss_equals_case (p->token.string, ss_cstr ("!TO")))
+ {
+ macro_error (stack, p < end ? p : NULL,
+ _("Expected !TO in numerical !DO loop."));
+ return 0;
+ }
+ p++;
+
+ double last;
+ if (!macro_evaluate_number (&p, end - p, nesting_countdown,
+ segmenter_mode, macros, me, &next_stack,
+ vars, expand, &last))
+ return 0;
+
+ double by = 1.0;
+ if (p < end && p->token.type == T_MACRO_ID
+ && ss_equals_case (p->token.string, ss_cstr ("!BY")))
+ {
+ p++;
+ if (!macro_evaluate_number (&p, end - p, nesting_countdown,
+ segmenter_mode, macros, me, &next_stack,
+ vars, expand, &by))
+ return 0;
+
+ if (by == 0.0)
+ {
+ macro_error (stack, NULL, _("!BY value cannot be zero."));
+ return 0;
+ }
+ }
+
+ const struct macro_token *do_end = find_doend (stack, p, end);
+ if (!do_end)
+ return 0;
+ const struct macro_tokens inner = {
+ .mts = CONST_CAST (struct macro_token *, p),
+ .n = do_end - p
+ };
+
+ if ((by > 0 && first <= last) || (by < 0 && first >= last))
+ {
+ int i = 0;
+ for (double index = first;
+ by > 0 ? (index <= last) : (index >= last);
+ index += by)
+ {
+ if (i++ > miterate)
+ {
+ macro_error (stack, NULL,
+ _("Numerical !DO loop exceeded "
+ "maximum number of iterations %d. "
+ "(Use SET MITERATE to change the limit.)"),
+ miterate);
+ break;
+ }
+
+ char index_s[DBL_BUFSIZE_BOUND];
+ c_dtoastr (index_s, sizeof index_s, 0, 0, index);
+ string_map_replace_nocopy (vars, ss_xstrdup (var_name),
+ xstrdup (index_s));
+
+ bool break_ = false;
+ macro_expand (&inner, nesting_countdown, segmenter_mode, macros,
+ me, vars, &next_stack, expand, &break_, exp);
+ if (break_)
+ break;
+ }
+ }
+
+ return do_end - tokens + 1;
+ }
+ else
+ {
+ macro_error (stack, p < end ? p : NULL,
+ _("Expected `=' or !IN in !DO loop."));
+ return 0;
+ }
+}
+
+static void
+macro_expand (const struct macro_tokens *mts, int nesting_countdown,
+ enum segmenter_mode segmenter_mode,
+ const struct macro_set *macros,
+ const struct macro_expander *me, struct string_map *vars,
+ const struct macro_expansion_stack *stack,
+ bool *expand, bool *break_, struct macro_tokens *exp)
+{
+ if (nesting_countdown <= 0)
+ {
+ macro_error (stack, NULL, _("Maximum nesting level %d exceeded. "
+ "(Use SET MNEST to change the limit.)"),
+ settings_get_mnest ());
+ for (size_t i = 0; i < mts->n; i++)
+ macro_tokens_add (exp, &mts->mts[i]);
+ return;
+ }
+
+ struct string_map own_vars = STRING_MAP_INITIALIZER (own_vars);
+ if (!vars)
+ vars = &own_vars;
+
+ for (size_t i = 0; i < mts->n && (!break_ || !*break_); i++)
+ {
+ const struct macro_token *mt = &mts->mts[i];
+ const struct token *token = &mt->token;
+ if (token->type == T_MACRO_ID && me)
+ {
+ const struct macro_param *param = macro_find_parameter_by_name (
+ me->macro, token->string);
+ if (param)
+ {
+ const struct macro_tokens *arg = me->args[param - me->macro->params];
+ if (*expand && param->expand_arg)
+ macro_expand (arg, nesting_countdown, segmenter_mode,
+ macros, NULL, NULL,
+ &(struct macro_expansion_stack) {
+ .name = param->name,
+ .next = stack,
+ }, expand, break_, exp);
+ else
+ for (size_t i = 0; i < arg->n; i++)
+ macro_tokens_add (exp, &arg->mts[i]);
+ continue;
+ }
+
+ if (is_bang_star (mts->mts, mts->n, i))
+ {
+ for (size_t j = 0; j < me->macro->n_params; j++)
+ {
+ const struct macro_param *param = &me->macro->params[j];
+ if (!param->positional)
+ break;
+
+ const struct macro_tokens *arg = me->args[j];
+ if (*expand && param->expand_arg)
+ macro_expand (arg, nesting_countdown, segmenter_mode,
+ macros, NULL, NULL,
+ &(struct macro_expansion_stack) {
+ .name = "!*",
+ .next = stack,
+ }, expand, break_, exp);
+ else
+ for (size_t k = 0; k < arg->n; k++)
+ macro_tokens_add (exp, &arg->mts[k]);
+ }
+ i++;
+ continue;
+ }
+
+ size_t n = macro_expand_if (&mts->mts[i], mts->n - i,
+ nesting_countdown, segmenter_mode,
+ macros, me, stack,
+ vars, expand, break_, exp);
+ if (n > 0)
+ {
+ i += n - 1;
+ continue;
+ }
+ }
+
+ if (token->type == T_MACRO_ID && vars)
+ {
+ const char *value = string_map_find__ (vars, token->string.string,
+ token->string.length);
+ if (value)
+ {
+ macro_tokens_from_string__ (exp, ss_cstr (value), segmenter_mode,
+ stack);
+ continue;
+ }
+ }
+
+ if (*expand)
+ {
+ struct macro_expander *subme;
+ int retval = macro_expander_create (macros, token, &subme);
+ for (size_t j = 1; !retval; j++)
+ {
+ const struct macro_token endcmd = { .token = { .type = T_ENDCMD } };
+ retval = macro_expander_add (
+ subme, i + j < mts->n ? &mts->mts[i + j] : &endcmd);
+ }
+ if (retval > 0)
+ {
+ i += retval - 1;
+ macro_expand (&subme->macro->body, nesting_countdown - 1,
+ segmenter_mode, macros, subme, NULL,
+ &(struct macro_expansion_stack) {
+ .name = subme->macro->name,
+ .file_name = subme->macro->file_name,
+ .first_line = subme->macro->first_line,
+ .last_line = subme->macro->last_line,
+ .next = stack,
+ }, expand, break_, exp);
+ macro_expander_destroy (subme);
+ continue;
+ }
+
+ macro_expander_destroy (subme);
+ }
+
+ if (token->type != T_MACRO_ID)
+ {
+ macro_tokens_add (exp, mt);
+ continue;
+ }
+
+ if (ss_equals_case (token->string, ss_cstr ("!break")))
+ {
+ if (!break_)
+ macro_error (stack, mt, _("!BREAK outside !DO."));
+ else
+ {
+ *break_ = true;
+ break;
+ }
+ }
+
+ struct parse_macro_function_ctx ctx = {
+ .input = &mts->mts[i],
+ .n_input = mts->n - i,
+ .nesting_countdown = nesting_countdown,
+ .segmenter_mode = segmenter_mode,
+ .macros = macros,
+ .me = me,
+ .stack = stack,
+ .vars = vars,
+ .expand = expand,
+ };
+ struct string function_output = DS_EMPTY_INITIALIZER;
+ size_t function_consumed;
+ if (expand_macro_function (&ctx, &function_output, &function_consumed))
+ {
+ i += function_consumed - 1;
+
+ macro_tokens_from_string__ (exp, function_output.ss, segmenter_mode,
+ stack);
+ ds_destroy (&function_output);
+
+ continue;
+ }
+
+ size_t n = macro_parse_let (&mts->mts[i], mts->n - i,
+ nesting_countdown, segmenter_mode,
+ macros, me, stack, vars, expand);
+ if (n > 0)
+ {
+ i += n - 1;
+ continue;
+ }
+
+ n = macro_expand_do (&mts->mts[i], mts->n - i,
+ nesting_countdown, segmenter_mode,
+ macros, me, stack, vars, expand, exp);
+ if (n > 0)
+ {
+ i += n - 1;
+ continue;
+ }
+
+ if (ss_equals_case (token->string, ss_cstr ("!onexpand")))
+ *expand = true;
+ else if (ss_equals_case (token->string, ss_cstr ("!offexpand")))
+ *expand = false;
+ else
+ macro_tokens_add (exp, mt);
+ }
+ if (vars == &own_vars)
+ string_map_destroy (&own_vars);
+}
+
+void
+macro_expander_get_expansion (struct macro_expander *me,
+ enum segmenter_mode segmenter_mode,
+ struct macro_tokens *exp)
+{
+ assert (me->state == ME_FINISHED);
+ bool expand = true;
+ struct macro_expansion_stack stack = {
+ .name = me->macro->name,
+ .file_name = me->macro->file_name,
+ .first_line = me->macro->first_line,
+ .last_line = me->macro->last_line,
+ };
+ macro_expand (&me->macro->body, settings_get_mnest (), segmenter_mode,
+ me->macros, me, NULL, &stack, &expand, NULL, exp);
+}
+