-
Although Composer is quite user-friendly and can modify multiple files to generate code as well as provide explanations, when it comes to Chinese characters, symbols like “��” often appear to replace the Chinese characters instead.
-
For the code generated by Composer, I often find that when my prompt words clearly indicate modifying function B in the current script, Composer will modify the code in other functions while modifying function B and will also generate symbols like " ��". This is completely unnecessary and shows that it doesn’t fully understand natural language.
-
In the new version, Composer allows viewing the history records. I’ve noticed that when there are a bit more history records in Composer, when I open the Cursor program, the screen will go black, show a spinning circle, and become unresponsive. After about 10 seconds, a prompt will appear: “The application is not responding. Do you want to restart Cursor, close it, or continue waiting?” After choosing to restart Cursor, another prompt will show up again after about 10 seconds: “The application is not responding. Do you want to restart Cursor, close it, or continue waiting?” Only by clicking “Continue waiting” can the main editing screen be displayed normally after about 5 seconds. At first, I thought this was an issue with the version. Reinstalling the old version didn’t solve the problem. After uninstalling the current version and reinstalling the old version again, when I opened Cursor for the second time, I encountered the lagging problem again. Later, I accidentally found that deleting Composer’s history records could solve this lagging issue.
-
When there are more history records in Composer, the content of the responses sometimes gets messed up. Sometimes the responses are also very slow, and clicking on the content replied by Composer or the content of the scripts modified by Composer is also very laggy.
-
Composer can’t accurately modify the code according to the prompt words. I’ve found that sometimes my prompt words ask Composer to modify the code to fix bugs, but in order to fix the bugs by modifying the code, Composer will delete all the code in this function or this section of code that has nothing to do with the prompt words. So I’m really confused. What’s the point of Composer fixing the bugs when my main business logic code has been deleted?
-
It lacks logical understanding and judgment, especially when multiple scripts are related. For example, when writing a Python Flask web application, both the front-end code and the back-end code need to be written simultaneously and the parameter requests need to match. But most of the time, Composer can’t find logical errors or errors where parameters and variables don’t match between the front end and the back end by reading the original code. Unless I give it clear error logs or tell it where the problem is, I often have to check the front-end and back-end matching issues of code parameters and variables line by line by myself, find the errors, tell Composer about them, and then let Composer help me fix the problems.
-
Regarding the issue of bracket pair colorization, I’ve checked and found that it’s enabled by default. However, when there are multiple brackets in the code, when I select an opening bracket, the corresponding closing bracket isn’t marked, and it’s impossible to tell them apart at a glance. PyCharm offers a much better experience in this regard. I’m not sure if I didn’t set it up properly. I saw that the status of “Bracket Pair Colorization” in the settings is “enabled”.
In the settings, all the artificial intelligence rules are only for Chat and don’t support Composer. I think even if the prompt words aren’t accurate enough and Composer can’t understand them, it shouldn’t modify or delete the code or even functions that have nothing to do with the prompt words.
These are the issues for now. I hope to get a reply or have them solved.