Fix issue #907 and #888, always filter candidates at Python side, avo…

…id acm menu not show completion candidates when candidates return by LSP server much bigger than the limit of acm-backend-lsp-candidates-max-number

README.md

@@ -338,7 +338,7 @@ lsp-bridge provides support for more than two language servers for many language
 - `acm-backend-yas-candidates-number`: yasnippet display number， 2 by default
 - `acm-backend-citre-keyword-complete`: Completion is performed according to the keywords of each mode defined by `acm-backend-citre-keywords-alist`, which takes effect only after citre is enabled.
 - `acm-backend-search-sdcv-words-dictionary`: StarDict dictionary for word completion, default is `kdic-ec-11w`, you can replace it with StarDict dictionary path, example, if you have dictionary `/usr/share/stardict/dic/stardict-oxford-gb-formated-2.4.2/oxford-gb-formated.ifo`, you need set this value to `/usr/share/stardict/dic/stardict-oxford-gb-formated-2.4.2/oxford-gb-formated`, not include `.ifo` extension.
-- `acm-backend-lsp-match-mode`: The filtering mode for candidate words in LSP backend, there are three options: "normal", "prefix", "prefixCaseSensitive", and "fuzzy". By default it is normal. It does not filter the candidate words returned by LSP Server
+- `acm-backend-lsp-match-mode`: The filtering mode for candidates in LSP backend, there are three options: "prefix", "prefixCaseSensitive", and "fuzzy". By default it is "fuzzy"
 - `acm-enable-preview`: enable Tab-and-Go completion, commands like acm-select-* will select and preview other candidate and further input will then commit this candidate, disable by default
 
 ## Add support for new language?

README.zh-CN.md

@@ -336,7 +336,7 @@ lsp-bridge 针对许多语言都提供 2 个以上的语言服务器支持， 
 - `acm-backend-yas-candidates-number`: yasnippet 显示个数， 默认 2 个
 - `acm-backend-citre-keyword-complete`: 根据`acm-backend-citre-keywords-alist`定义的各个模式的关键字进行补全， 需要使能 citre 后才生效
 - `acm-backend-search-sdcv-words-dictionary`: 用于单词补全的 StarDict 词典， 默认是 `kdic-ec-11w`, 可以自定义为其他 StarDict 词典， 如果你的系统存在词典 `/usr/share/stardict/dic/stardict-oxford-gb-formated-2.4.2/oxford-gb-formated.ifo`, 你需要设置这个选项为 `/usr/share/stardict/dic/stardict-oxford-gb-formated-2.4.2/oxford-gb-formated`, 不需要包括 `.ifo` 扩展
-- `acm-backend-lsp-match-mode`: LSP 后端候选词过滤模式， 有 "normal", "prefix", "prefixCaseSensitive", "fuzzy" 三个选项， 默认是 "normal", 不对 LSP Server 返回候选词进行过滤
+- `acm-backend-lsp-match-mode`: LSP 后端候选词过滤模式， 有 "prefix", "prefixCaseSensitive", "fuzzy" 三个选项， 默认是 "fuzzy"
 - `acm-enable-preview`: 开启 Tab-and-Go completion， 当改变当前候选时， 可以预览候选， 并且后续输入会选择预览候选， 默认关闭
 
 ## 添加新的编程语言支持?

acm/acm-backend-lsp.el

@@ -108,15 +108,18 @@
   :type 'boolean
   :group 'acm-backend-lsp)
 
-(defcustom acm-backend-lsp-match-mode "normal"
+(defcustom acm-backend-lsp-match-mode "fuzzy"
   "The match mode to filter completion candidates.
 
-normal: don't filter candidates.
-prefix: filter candidates with input prefix, note such as C++, after `std::', candidate's prefix is not `::'
-prefixCaseSensitive: filter candidates with input prefix, and case sensitive
-fuzzy: fitler candidates with fuzzy algorithm
+`prefix': filter candidates with input prefix, note such as C++, after `std::', candidate's prefix is not `::'
+`prefixCaseSensitive': filter candidates with input prefix, and case sensitive
+`fuzzy': fitler candidates with fuzzy algorithm
 
-Recommand use `normal' that follow LSP server response, emacser's behavior typically does not adapt to LSP protocol."
+lsp-bridge still will use `fuzzy' algorithm filter candidates if value is not `prefix' `prefixCaseSensitive' or `fuzzy'.
+
+The lsp-bridge will continuously filter candidates on the Python side.
+If not filter and the value of `acm-backend-lsp-candidates-max-number' is far smaller than the number of candidates returned by the LSP server,
+it will cause the lsp-bridge to always send the previous batch of candidates which do not match the users input."
   :type 'string
   :group 'acm-backend-lsp)
 

core/handler/completion.py

@@ -126,20 +126,23 @@ def process_response(self, response: dict) -> None:
                 except:
                     pass
 
-                # Try to drop current candidate if it match user rule.
+                # The lsp-bridge will continuously filter candidates on the Python side.
+                # If not filter and the value of `acm-backend-lsp-candidates-max-number' is far smaller
+                # than the number of candidates returned by the LSP server,
+                # it will cause the lsp-bridge to always send the previous batch of candidates
+                # which do not match the users input.
                 if match_mode == "prefix":
                     if not string_match(label.lower(), self.prefix.lower(), fuzzy=False):
                         continue
                 elif match_mode == "prefixCaseSensitive":
                     if not string_match(label, self.prefix, fuzzy=False):
                         continue
-                elif match_mode == "fuzzy":
+                else:
                     if not string_match(label.lower(), self.prefix.lower(), fuzzy=True):
                         continue
 
                 annotation = kind if kind != "" else detail
 
-
                 # The key keyword combines the values ​​of 'label' and 'detail'
                 # to handle different libraries provide the same function.
                 key = f"{label}_{detail}"