From 5cd404bdc6b0d56327bfc98fdec7b817f850a6cb Mon Sep 17 00:00:00 2001 From: Yuan Fu Date: Tue, 16 Apr 2024 23:46:39 -0700 Subject: [PATCH] New function treesit-parser-changed-ranges - Add a new field last_changed_ranges to tree-sitter parser object. - Add a new function treesit-parser-changed-ranges * doc/lispref/parsing.texi (Using Parser): Add the function in tree-sitter manual. * src/treesit.c (treesit_get_changed_ranges): New function, refactored out of treesit_call_after_change_functions. (treesit_call_after_change_functions): Pull out treesit_get_changed_ranges. (treesit_ensure_parsed): Save the changed ranges to the parser object. (make_treesit_parser): Initialize the new parser field last_changed_ranges. (Ftreesit_parser_changed_ranges): New function. (Qtreesit_unparsed_edits): New error. * src/treesit.h (Lisp_TS_Parser): New field. (cherry picked from commit 996b9576713f9d63ea7ff7e9630a15cb0a0214eb) --- doc/lispref/parsing.texi | 20 ++++++++++++++++ etc/NEWS | 3 +++ src/treesit.c | 49 ++++++++++++++++++++++++++++++++++++---- src/treesit.h | 3 +++ 4 files changed, 71 insertions(+), 4 deletions(-) diff --git a/doc/lispref/parsing.texi b/doc/lispref/parsing.texi index 5fd1eaaa57e..65672997bda 100644 --- a/doc/lispref/parsing.texi +++ b/doc/lispref/parsing.texi @@ -539,6 +539,26 @@ symbol, rather than a lambda function. This function returns the list of @var{parser}'s notifier functions. @end defun +Sometimes a user might want to synchronously get the changed ranges of +the last reparse, and @code{treesit-parser-changed-ranges} is just for +it. This function basically returns the @var{ranges} that the notifier +functions were passed. + +@defun treesit-parser-changed-ranges parser &optional quiet +This function returns the ranges that has been changed since last +reparse. It returns a list of cons cells of the form +@w{@code{(@var{start} . @var{end})}}, where @var{start} and @var{end} +mark the start and the end positions of a range. + +This function should almost always be called immediately after +reparsing. If it's called when there are new buffer edits that hasn't +been reparsed, Emacs signals @code{treesit-unparsed-edits}, unless +@var{quiet} is non-nil. + +Calling this function multiple times consecutively doesn't change its +return value; it always returns the ranges affected by the last reparse. +@end defun + @node Retrieving Nodes @section Retrieving Nodes @cindex retrieve node, tree-sitter diff --git a/etc/NEWS b/etc/NEWS index b04d2f4be34..549bbe01926 100644 --- a/etc/NEWS +++ b/etc/NEWS @@ -2688,6 +2688,9 @@ only return parsers for that language. If TAG is given, only return parsers with that tag. Note that passing nil as tag doesn't mean return all parsers, but rather "all parsers with no tags". ++++ +*** New function 'treesit-parser-changed-ranges' which returns buffer regions that are affected by the last buffer edits + * Changes in Emacs 30.1 on Non-Free Operating Systems diff --git a/src/treesit.c b/src/treesit.c index d86ab501187..76354361284 100644 --- a/src/treesit.c +++ b/src/treesit.c @@ -1017,9 +1017,8 @@ treesit_check_buffer_size (struct buffer *buffer) static Lisp_Object treesit_make_ranges (const TSRange *, uint32_t, struct buffer *); -static void -treesit_call_after_change_functions (TSTree *old_tree, TSTree *new_tree, - Lisp_Object parser) +static Lisp_Object +treesit_get_changed_ranges (TSTree *old_tree, TSTree *new_tree, Lisp_Object parser) { /* If the old_tree is NULL, meaning this is the first parse, the changed range is the whole buffer. */ @@ -1039,7 +1038,13 @@ treesit_call_after_change_functions (TSTree *old_tree, TSTree *new_tree, lisp_ranges = Fcons (Fcons (Fpoint_min (), Fpoint_max ()), Qnil); set_buffer_internal (oldbuf); } + return lisp_ranges; +} +static void +treesit_call_after_change_functions (Lisp_Object lisp_ranges, + Lisp_Object parser) +{ specpdl_ref count = SPECPDL_INDEX (); /* let's trust the after change functions and not clone a new ranges @@ -1091,13 +1096,17 @@ treesit_ensure_parsed (Lisp_Object parser) XTS_PARSER (parser)->tree = new_tree; XTS_PARSER (parser)->need_reparse = false; + Lisp_Object changed_ranges; + changed_ranges = treesit_get_changed_ranges (tree, new_tree, parser); + XTS_PARSER (parser)->last_changed_ranges = changed_ranges; + /* After-change functions should run at the very end, most crucially after need_reparse is set to false, this way if the function calls some tree-sitter function which invokes treesit_ensure_parsed again, it returns early and do not recursively call the after change functions again. (ref:notifier-inside-ensure-parsed) */ - treesit_call_after_change_functions (tree, new_tree, parser); + treesit_call_after_change_functions (changed_ranges, parser); ts_tree_delete (tree); } @@ -1171,6 +1180,7 @@ make_treesit_parser (Lisp_Object buffer, TSParser *parser, lisp_parser->after_change_functions = Qnil; lisp_parser->tag = tag; lisp_parser->last_set_ranges = Qnil; + lisp_parser->last_changed_ranges = Qnil; lisp_parser->buffer = buffer; lisp_parser->parser = parser; lisp_parser->tree = tree; @@ -1818,6 +1828,32 @@ positions. PARSER is the parser issuing the notification. */) return Qnil; } +DEFUN ("treesit-parser-changed-ranges", Ftreesit_parser_changed_ranges, + Streesit_parser_changed_ranges, + 1, 2, 0, + doc: /* Return the buffer regions affected by the last reparse of PARSER. + +Returns a list of cons (BEG . END), where each cons represents a region +in which the buffer content was affected by the last reparse. + +This function should almost always be called immediately after +reparsing. If it's called when there are new buffer edits that hasn't +been reparsed, Emacs signals `treesit-unparsed-edits', unless QUIET is +non-nil. + +Calling this function multiple times consecutively doesn't change its +return value; it always returns the ranges affected by the last +reparse. */) + (Lisp_Object parser, Lisp_Object quiet) +{ + treesit_check_parser (parser); + + if (XTS_PARSER (parser)->need_reparse && NILP (quiet)) + xsignal1 (Qtreesit_unparsed_edits, parser); + + return XTS_PARSER (parser)->last_changed_ranges; +} + /*** Node API */ @@ -4010,6 +4046,7 @@ syms_of_treesit (void) DEFSYM (Qtreesit_query_error, "treesit-query-error"); DEFSYM (Qtreesit_parse_error, "treesit-parse-error"); DEFSYM (Qtreesit_range_invalid, "treesit-range-invalid"); + DEFSYM (Qtreesit_unparsed_edits, "treesit-unparsed_edits"); DEFSYM (Qtreesit_buffer_too_large, "treesit-buffer-too-large"); DEFSYM (Qtreesit_load_language_error, @@ -4038,6 +4075,8 @@ syms_of_treesit (void) define_error (Qtreesit_range_invalid, "RANGES are invalid: they have to be ordered and should not overlap", Qtreesit_error); + define_error (Qtreesit_unparsed_edits, "There are unparsed edits in the buffer", + Qtreesit_error); define_error (Qtreesit_buffer_too_large, "Buffer too large (> 4GiB)", Qtreesit_error); define_error (Qtreesit_load_language_error, @@ -4178,6 +4217,8 @@ the symbol of that THING. For example, (or sexp sentence). */); defsubr (&Streesit_parser_add_notifier); defsubr (&Streesit_parser_remove_notifier); + defsubr (&Streesit_parser_changed_ranges); + defsubr (&Streesit_node_type); defsubr (&Streesit_node_start); defsubr (&Streesit_node_end); diff --git a/src/treesit.h b/src/treesit.h index bb81bf0e2b3..aa71933fe8d 100644 --- a/src/treesit.h +++ b/src/treesit.h @@ -49,6 +49,9 @@ struct Lisp_TS_Parser ranges the users wants to set, and avoid reparse if the new ranges is the same as the last set one. */ Lisp_Object last_set_ranges; + /* The range of buffer content that was affected by the last + re-parse. */ + Lisp_Object last_changed_ranges; /* The buffer associated with this parser. */ Lisp_Object buffer; /* The pointer to the tree-sitter parser. Never NULL. */ -- 2.39.5