suwa auto checkin

Background

Recently I created an account in suwa, with a small quota of data. But everyday checkin will additionally give me several hundred MB for free.

So I like to do it automatically. It took me a weekend to wrap up this with playwright. Playwright is awesome, although I am new to it but easy to write with it. Compared to puppeteer, it is obvious that playwright is much better and well-documented.

During the login, suwa requires to move a slider for verification. I choose to do it in pure front-end without installing heavy opencv related libs in backend. So the code will rely on opencv.js, in this experiment, opencv 4.5.5 works better than 5.x dev version.

How to run

• make sure in your home directory, there is a file named suwa_credential.txt, with user name in first line and password in second line.

• run in headless mode:

npx playwright test

or headed mode for debug, with xming setup correctly if you have no GUI in the server where it runs:

npx playwright test --headed

Of course, need to put it into crontab if run it periodically.

Screenshot

This is one of screenshots, the canvas for edges is not necessary, only show it for visual looks.

Knowledge

Want to share what I learn from this test.

• selector is the key of playwright, it is mostly like css selector. page.locator("button:has(span) .MuiButton-label") represents AND relation, while page.locator("button:has-text('abc'),button:has-text('xyz') stands for OR, waiting for appearance of button abc or xyz. So far I can only write simple selector syntax. One trick is writing like await page.locator("button:has(span)").click(), this will run into error according to strickness, the error msg shows the elements resolved and corresponding matching syntax, but it only show maximum 10 entries. The easier way is to list all buttons, then loop to find expected one.

• evaluate in playwright runs the pageFunction in browser context, not in node context. This bothers me much at the very beginning. With the imported several opencv js bindings, like this and this, use cv in pageFunction, in runtime xxx is not defined is seen. So official opencv.js is passed as a string parameter into pageFunction, it is size of 8M+, currently it is fine until it becomes huge in the future :). You may notice that inside evaluate pageFunction, there are no types, considering the types in aforesaid third-party wrappers may not be consistent with what I actually use in pageFunction.

• The same, console.log in evaluate pageFunction could not print well to node console, unless it is run in headed mode and check it in browser dev console. Use page.on("console",...) is better but still could not get the object printed, instead, the output is something like src is {}.

Maybe I miss something so that it works like this, but I hope this could help other guys who encounter the similar issues.