From 3ae2a5c2f666bae94f756407e262ebce7b5f5ef2 Mon Sep 17 00:00:00 2001 From: Maxime Coste Date: Sat, 7 Nov 2020 10:19:11 +1100 Subject: [PATCH] Fix performance issue with word completion When pasting many words with we can end-up with a huge concatenated word and many selections, the previous code ended up iterating from each selection cursor to that word start and end to find the word under the cursor. This could lead to performance issue as each selection would trigger iteration on that huge word. This is unnecessary as word completion has a word length limit, so we now take it into account to avoid iterating to far from the cursor position. --- src/insert_completer.cc | 14 ++++++++++---- src/word_db.cc | 4 +--- src/word_db.hh | 2 ++ 3 files changed, 13 insertions(+), 7 deletions(-) diff --git a/src/insert_completer.cc b/src/insert_completer.cc index 03ce45a5..ca6b08a1 100644 --- a/src/insert_completer.cc +++ b/src/insert_completer.cc @@ -86,9 +86,12 @@ InsertCompletion complete_word(const SelectionList& sels, HashMap sel_word_counts; for (int i = 0; i < sels.size(); ++i) { + int len = 0; + auto is_short_enough_word = [&] (Codepoint c) { return len++ < WordDB::max_word_len && is_word_pred(c); }; + Utf8It end{buffer.iterator_at(sels[i].cursor()), buffer}; Utf8It begin = end-1; - if (not skip_while_reverse(begin, buffer.begin(), is_word_pred) and + if (not skip_while_reverse(begin, buffer.begin(), is_short_enough_word) and begin < end) // (begin might == end if end == buffer.begin()) ++begin; @@ -98,10 +101,13 @@ InsertCompletion complete_word(const SelectionList& sels, prefix = buffer.substr(word_begin, end.base().coord()); } - skip_while(end, buffer.end(), is_word_pred); + skip_while(end, buffer.end(), is_short_enough_word); - StringView word = buffer.substr(begin.base().coord(), end.base().coord()); - ++sel_word_counts[word]; + if (len <= WordDB::max_word_len) + { + StringView word = buffer.substr(begin.base().coord(), end.base().coord()); + ++sel_word_counts[word]; + } } struct RankedMatchAndBuffer : RankedMatch diff --git a/src/word_db.cc b/src/word_db.cc index 17db080b..ed0c7d99 100644 --- a/src/word_db.cc +++ b/src/word_db.cc @@ -21,8 +21,6 @@ WordDB& get_word_db(const Buffer& buffer) struct WordSplitter { - static constexpr CharCount max_word_len = 50; - struct Iterator { Iterator(const char* begin, const WordSplitter& splitter) @@ -48,7 +46,7 @@ struct WordSplitter utf8::to_next(m_word_end, end); ++word_len; } - if (m_word_begin == end or word_len < max_word_len) + if (m_word_begin == end or word_len < WordDB::max_word_len) break; } diff --git a/src/word_db.hh b/src/word_db.hh index dab65d4c..ac7b8c01 100644 --- a/src/word_db.hh +++ b/src/word_db.hh @@ -18,6 +18,8 @@ class Buffer; class WordDB : public OptionManagerWatcher { public: + static constexpr CharCount max_word_len = 50; + WordDB(const Buffer& buffer); ~WordDB(); WordDB(const WordDB&) = delete;