/* SPDX-License-Identifier: GPL-2.0 */

#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#include <stdbool.h>
#include <limits.h>
#include <string.h>
#include <errno.h>
#include <inttypes.h>
#include <fnmatch.h>
#include <regex.h>
#include <assert.h>
#include <lopsub.h>

#include "err.h"

/* Opaque, only known to ast.c. Passed to the generated txp_yyparse(). */
struct txp_context;

/*
 * Since we use a reentrant lexer, all functions generated by flex(1)
 * receive an additional argument of this type.
 */
typedef void *txp_yyscan_t;

/* Parsed regex pattern. */
struct txp_re_pattern {
	regex_t preg; /* Pre-compiled regex. */
	unsigned flags; /* Subset of the cflags described in regex(3). */
};

/*
 * The possible values of a node in the abstract syntax tree (AST).
 *
 * Constant semantic values (string literals, numeric constants and regex
 * patterns which are part of the tag expression) are determined during
 * txp_init() while values which depend on the epigram (tags, number of lines,
 * etc.) are determined during txp_eval_row().
 *
 * This union, and the txp_ast_node structure below are used extensively in
 * txp.y. However, both need to be public because the lexer must be able to
 * create AST nodes for the constant semantic values.
 */
union txp_semantic_value {
	bool boolval; /* Comparators, =~ and =|. */
	char *strval; /* String literals, tags, path. */
	int64_t intval; /* Constants, bitrate, frequency, etc. */
	struct txp_re_pattern re_pattern; /*< Right-hand side operand of =~. */
};

/*
 * A node is either interior or a leaf node. Interior nodes have at least one
 * child while leaf nodes have a semantic value and no children.
 *
 * Examples: (a) STRING_LITERAL has a semantic value (the unescaped string
 * literal) and no children, (b) NEG (unary minus) has no semantic value but
 * one child (the numeric expression that is to be negated), (c) LESS_OR_EQUAL
 * has no semantic value and two children (the two numeric expressions being
 * compared).
 */
struct txp_ast_node {
	/* Corresponds to a token type, for example LESS_OR_EQUAL. */
	int id;
	union {
		/* Pointers to the child nodes (interior nodes only). */
		struct txp_ast_node **children;
		/* Leaf nodes only. */
		union txp_semantic_value sv;
	};
	/*
	 * The number of children is implicitly given by the id, but we include
	 * it here to avoid having to maintain a lookup table. The AST is
	 * usually small, so we can afford to waste a byte per node.
	 */
	uint8_t num_children;
};

enum loglevels {LOGLEVELS, NUM_LOGLEVELS};
#define DEBUG_LOG(f,...) txp_log(LL_DEBUG, "%s: " f, __FUNCTION__, ## __VA_ARGS__)
#define INFO_LOG(f,...) txp_log(LL_INFO, "%s: " f, __FUNCTION__, ## __VA_ARGS__)
#define NOTICE_LOG(f,...) txp_log(LL_NOTICE, "%s: " f, __FUNCTION__, ## __VA_ARGS__)
#define WARNING_LOG(f,...) txp_log(LL_WARNING, "%s: " f, __FUNCTION__, ##  __VA_ARGS__)
#define ERROR_LOG(f,...) txp_log(LL_ERROR, "%s: " f, __FUNCTION__, ## __VA_ARGS__)
#define CRIT_LOG(f,...) txp_log(LL_CRIT, "%s: " f, __FUNCTION__, ## __VA_ARGS__)
#define EMERG_LOG(f,...) txp_log(LL_EMERG, "%s: " f, __FUNCTION__, ## __VA_ARGS__)

/* tfortune.c */

void txp_log(int ll, const char* fmt,...);

/* Called from both the lexer and the parser. */
__attribute__ ((format (printf, 3, 4)))
void txp_parse_error(int line, struct txp_context *ctx, const char *fmt, ...);

/* Helper functions for the lexer. */
unsigned parse_quoted_string(const char *src, const char quote_chars[2],
		char **result);
int txp_parse_regex_pattern(const char *src, struct txp_re_pattern *result);

/* ast.c */
struct txp_ast_node *ast_node_new_unary(int id, struct txp_ast_node *child);
struct txp_ast_node *ast_node_new_binary(int id, struct txp_ast_node *left,
		struct txp_ast_node *right);

/*
 * Allocate a new leaf node for the abstract syntax tree.
 *
 * This returns a pointer to a node whose ->num_children field is initialized
 * to zero. The ->id field is initialized with the given id.  The caller is
 * expected to initialize the ->sv field.
 */
struct txp_ast_node *txp_new_ast_leaf_node(int id);

/*
 * Evaluate an abstract syntax tree, starting at the root node.
 *
 * The root node argument should be the pointer that was returned from an
 * earlier call to txp_init() via the context pointer. The context contains the
 * information about the epigram.
 *
 * Returns true if the AST evaluates to true, a non-empty string, or a non-zero
 * number, false otherwise.
 */
bool txp_eval_ast(struct txp_ast_node *root, struct txp_context *ctx);

/*
 * Deallocate an abstract syntax tree.
 *
 * This frees the memory occupied by the nodes of the AST, the child pointers
 * of the internal nodes and the (constant) semantic values of the leaf nodes
 * (string literals and pre-compiled regular expressions).
 */
void txp_free_ast(struct txp_ast_node *root);


/* util.c */
int atoi64(const char *str, int64_t *value);
unsigned xvasprintf(char **result, const char *fmt, va_list ap);
unsigned xasprintf(char **result, const char *fmt, ...);
void *xmalloc(size_t size);
void *xcalloc(size_t size);
int xregcomp(regex_t *preg, const char *regex, int cflags);

/* txp.c */

/* txp.y. */