程式撰寫者做開發時常會使用到API，但其中的字串參數卻只能在執行時期才能被檢測到是否與正確格式相符。為了節省重複嘗試與查找資料的時間，我們提出一個針對字串參數問題的靜態分析檢查器來輔助使用者撰寫及重用程式碼。既存研究將字面值常數的抽象語法樹一起傳入檢查函式，或是以字面值型態與型別提示搭配第三方工具做檢測。其中會有兩個問題，(一)使用變數做參數控制會出現錯誤的分析訊息，(二)檢查資訊與程式撰寫者使用版本可能不同的隱憂。我們提出以爬蟲程式即時爬取線上指定版本，確保與使用者安裝的版本資料一致，用剖析抽象語法樹的方式定位字串參數的位置做檢查。對於字串變數經過條件式分支產生的多可能性字串，藉由變數數值表記錄賦值行為一併做檢查。另外將錯誤字串與可使用的字串列表做字串相似度比對，在錯誤報告上給予更精準的修改建議。此概念以pandas、TensorFlow為實作範例。;Programmers often use APIs when developing, but the incorrect string parameters in it can only be detected at run-time. In order to reduce the time of repeated execution and searching for documentation, we propose a static analysis checker for string parameters to assist programmers in writing and reusing code. There are two problems in related researches. One is wrong analysis reports at some situations. Another problem is the gap between the latest documentation and API version that programmers used. We propose a tool, StrChecker, to crawl the online documentation to ensure that it is consistent with the version installed by the user. And, we parse the abstract syntax tree to locate the position of the string parameters. For the multi-possibility strings generated by the conditional branch, we trace the assign statements to record all possible strings. In addition, we provide more accurate suggestions for modification on the report. StrChecker is compatible with most libraries, we implement this tool on pandas and TensorFlow.
