[Question]: Is playwhright-go thread-safe if multiple instances are created in separate go routines?

[martingrzzler]

I'm using this awesome library for web scraping but when running browsers in parallel I get random crashes.
So this might be because the library is not thread-safe. In the following example I create two different instances of Chrome with playwhright.Run. Is this save to do?

package main

import (
	"sync"

	"github.com/crawler/internal/crawl"
)

func main() {

	wg := sync.WaitGroup{}

	wg.Add(1)
	go func() {
		defer wg.Done()
                // creates a new instances with playwhright.Run
		b1, err := crawl.LauchBrowser(crawl.LaunchOptions{
			Headless: false,
		})
		if err != nil {
			panic("failed to launch browser" + err.Error())
		}
		for i := 0; i < 10; i++ {
			p, err := crawl.NewPage(b1)
			if err != nil {
				panic("failed to create new page" + err.Error())
			}

			_, err = crawl.ScrapeHomePage(p, "https://google.com")
			if err != nil {
				panic("failed to scrape home page" + err.Error())
			}

			err = p.Context().Close()
			if err != nil {
				panic("failed to close context" + err.Error())
			}
		}

	}()

	wg.Add(1)
	go func() {
		b2, err := crawl.LauchBrowser(crawl.LaunchOptions{
			Headless: false,
		})
		if err != nil {
			panic("failed to launch browser" + err.Error())
		}
		defer wg.Done()

		for i := 0; i < 10; i++ {
			p, err := crawl.NewPage(b2)
			if err != nil {
				panic("failed to create new page" + err.Error())
			}

			_, err = crawl.ScrapeHomePage(p, "https://example.com")
			if err != nil {
				panic("failed to scrape home page" + err.Error())
			}

			err = p.Context().Close()
			if err != nil {
				panic("failed to close context" + err.Error())
			}

		}
	}()

	wg.Wait()
}

[canstand]

What you should do is create a new BrowserContext per goroutine, instead of launching multiple browsers.

[martingrzzler]

@canstand So using multiple BrowserContext in each goroutine is safe then?
I'm asking because of this comment: #119 (comment)

[canstand]

It can be considered safe for your use case. BrowserContext is isolated from each other, which is equivalent to opening multiple browser windows and each has its own temporary data directory and profile.

You need to be careful to avoid deadlocks like #481 and #391 .