mirror of
https://github.com/rkd77/elinks.git
synced 2024-11-04 08:17:17 -05:00
69 lines
3.0 KiB
C
69 lines
3.0 KiB
C
|
#ifndef EL__ECMASCRIPT_ECMASCRIPT_H
|
||
|
#define EL__ECMASCRIPT_ECMASCRIPT_H
|
||
|
|
||
|
/* This is a trivial ECMAScript driver. All your base are belong to pasky. */
|
||
|
/* In the future you will get DOM, a complete ECMAScript interface and free
|
||
|
* plasm displays for everyone. */
|
||
|
|
||
|
#include "main/module.h"
|
||
|
#include "util/time.h"
|
||
|
|
||
|
struct string;
|
||
|
struct uri;
|
||
|
struct view_state;
|
||
|
|
||
|
struct ecmascript_interpreter {
|
||
|
struct view_state *vs;
|
||
|
void *backend_data;
|
||
|
time_t exec_start;
|
||
|
|
||
|
/* This is a cross-rerenderings accumulator of
|
||
|
* @document.onload_snippets (see its description for juicy details).
|
||
|
* They enter this list as they continue to appear there, and they
|
||
|
* never leave it (so that we can always find from where to look for
|
||
|
* any new snippets in document.onload_snippets). Instead, as we
|
||
|
* go through the list we maintain a pointer to the last processed
|
||
|
* entry. */
|
||
|
struct list_head onload_snippets; /* -> struct string_list_item */
|
||
|
struct string_list_item *current_onload_snippet;
|
||
|
|
||
|
/* ID of the {struct document} where those onload_snippets belong to.
|
||
|
* It is kept at 0 until it is definitively hard-attached to a given
|
||
|
* final document. Then if we suddenly appear with this structure upon
|
||
|
* a document with a different ID, we reset the state and start with a
|
||
|
* fresh one (normally, that does not happen since reloading sets
|
||
|
* ecmascript_fragile, but it can happen i.e. when the urrent document
|
||
|
* is reloaded in another tab and then you just cause the current tab
|
||
|
* to redraw. */
|
||
|
unsigned int onload_snippets_owner;
|
||
|
};
|
||
|
|
||
|
/* Why is the interpreter bound to {struct view_state} instead of {struct
|
||
|
* document}? That's easy, because the script won't raid just inside of the
|
||
|
* document, but it will also want to generate pop-up boxes, adjust form
|
||
|
* contents (which is doc_view-specific) etc. Of course the cons are that we
|
||
|
* need to wait with any javascript code execution until we get bound to the
|
||
|
* view_state through document_view - that means we are going to re-render the
|
||
|
* document if it contains a <script> area full of document.write()s. And why
|
||
|
* not bound the interpreter to {struct document_view} then? Because it is
|
||
|
* reset for each rerendering, and it sucks to do all the magic to preserve the
|
||
|
* interpreter over the rerenderings (we tried). */
|
||
|
|
||
|
struct ecmascript_interpreter *ecmascript_get_interpreter(struct view_state*vs);
|
||
|
void ecmascript_put_interpreter(struct ecmascript_interpreter *interpreter);
|
||
|
|
||
|
void ecmascript_reset_state(struct view_state *vs);
|
||
|
|
||
|
void ecmascript_eval(struct ecmascript_interpreter *interpreter, struct string *code);
|
||
|
unsigned char *ecmascript_eval_stringback(struct ecmascript_interpreter *interpreter, struct string *code);
|
||
|
/* Returns -1 if undefined. */
|
||
|
int ecmascript_eval_boolback(struct ecmascript_interpreter *interpreter, struct string *code);
|
||
|
|
||
|
/* Takes line with the syntax javascript:<ecmascript code>. Activated when user
|
||
|
* follows a link with this synstax. */
|
||
|
void ecmascript_protocol_handler(struct session *ses, struct uri *uri);
|
||
|
|
||
|
extern struct module ecmascript_module;
|
||
|
|
||
|
#endif
|