go-extractor/playwright.go

package extractor

import (
	"context"
	"errors"
	"fmt"
	"log/slog"
	"time"

	"github.com/playwright-community/playwright-go"
)

type playWrightBrowser struct {
	pw         *playwright.Playwright
	browser    playwright.Browser
	ctx        playwright.BrowserContext
	userAgent  string
	timeout    time.Duration
	cookieJar  CookieJar
	serverAddr string
}

var _ Browser = playWrightBrowser{}

type BrowserSelection string

var (
	ErrInvalidBrowserSelection = errors.New("invalid browser selection")
	ErrPageNotFound            = errors.New("page not found")
	ErrInvalidStatusCode       = errors.New("invalid status code")
)

const (
	BrowserChromium BrowserSelection = "chromium"
	BrowserFirefox  BrowserSelection = "firefox"
	BrowserWebKit   BrowserSelection = "webkit"
)

// DefaultFirefoxUserAgent is the user-agent string used for Firefox browser instances.
const DefaultFirefoxUserAgent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:147.0) Gecko/20100101 Firefox/147.0"

// DefaultChromiumUserAgent is the user-agent string used for Chromium browser instances.
const DefaultChromiumUserAgent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36"

// DefaultUserAgent is an alias for DefaultFirefoxUserAgent, retained for backward compatibility.
const DefaultUserAgent = DefaultFirefoxUserAgent

// Bool returns a pointer to the given bool value.
func Bool(v bool) *bool { return &v }

type Size struct {
	Width  int
	Height int
}
type BrowserOptions struct {
	UserAgent string                     // If empty, auto-selected based on Browser engine
	Browser   BrowserSelection // If unset defaults to Firefox.
	Timeout   *time.Duration             // If unset defaults to 30 seconds timeout.  If set to 0, no timeout

	// CookieJar will, if set, load all cookies from the cookie jar into the browser and save all cookies from the
	// browser into the cookie jar for each request.
	CookieJar

	ShowBrowser *bool // If nil, defaults to false (headless). Set to ptr to override.

	Dimensions Size
	DarkMode   bool

	// ServerAddress is the address of a Playwright server to connect to.
	// Defaults to the value of the environment variable PLAYWRIGHT_SERVER_ADDRESS.
	ServerAddress string

	// RequireServer will, if set, return an error if the connection to the
	// Playwright server fails instead of falling back to a local browser launch.
	RequireServer bool

	// UseLocalOnly will, if set, not connect to the Playwright server, and instead launch a local browser.
	UseLocalOnly bool

	// LaunchArgs are additional command-line arguments passed to the browser process.
	// For example: []string{"--disable-blink-features=AutomationControlled"}
	LaunchArgs []string

	// InitScripts are JavaScript snippets injected into every new browser context
	// before any page scripts run. Useful for overriding detectable properties like
	// navigator.webdriver.
	InitScripts []string

	// Stealth enables anti-bot-detection measures. When non-nil and true, common
	// evasions are applied automatically (launch args + init scripts). When nil,
	// defaults to true in NewBrowser / NewInteractiveBrowser.
	Stealth *bool
}

func sameSiteToPlaywright(s SameSite) *playwright.SameSiteAttribute {
	switch s {
	case SameSiteStrict:
		return playwright.SameSiteAttributeStrict
	case SameSiteLax:
		return playwright.SameSiteAttributeLax
	case SameSiteNone:
		return playwright.SameSiteAttributeNone
	default:
		return nil
	}
}

func playwrightSameSiteToSameSite(s *playwright.SameSiteAttribute) SameSite {
	if s == nil {
		return ""
	}
	switch *s {
	case *playwright.SameSiteAttributeStrict:
		return SameSiteStrict
	case *playwright.SameSiteAttributeLax:
		return SameSiteLax
	case *playwright.SameSiteAttributeNone:
		return SameSiteNone
	default:
		return ""
	}
}

func cookieToPlaywrightOptionalCookie(cookie Cookie) playwright.OptionalCookie {
	oc := playwright.OptionalCookie{
		Name:     cookie.Name,
		Value:    cookie.Value,
		Domain:   playwright.String(cookie.Host),
		Path:     playwright.String(cookie.Path),
		Expires:  playwright.Float(float64(cookie.Expires.Unix())),
		Secure:   playwright.Bool(cookie.Secure),
		HttpOnly: playwright.Bool(cookie.HttpOnly),
	}
	if cookie.SameSite != "" {
		oc.SameSite = sameSiteToPlaywright(cookie.SameSite)
	}
	return oc
}

func playwrightCookieToCookie(cookie playwright.Cookie) Cookie {
	return Cookie{
		Name:     cookie.Name,
		Value:    cookie.Value,
		Host:     cookie.Domain,
		Path:     cookie.Path,
		Expires:  time.Unix(int64(cookie.Expires), 0),
		Secure:   cookie.Secure,
		HttpOnly: cookie.HttpOnly,
		SameSite: playwrightSameSiteToSameSite(cookie.SameSite),
	}
}

func NewBrowser(ctx context.Context, opts ...BrowserOptions) (Browser, error) {
	var thirtySeconds = 30 * time.Second
	opt := mergeOptions(BrowserOptions{
		Browser: BrowserFirefox,
		Timeout: &thirtySeconds,
		Stealth: Bool(true),
		Dimensions: Size{
			Width:  1920,
			Height: 1080,
		},
	}, opts)

	if err := ctx.Err(); err != nil {
		return nil, err
	}

	type browserResult struct {
		browser Browser
		err     error
	}

	resultCh := make(chan browserResult, 1)

	go func() {
		res, err := initBrowser(opt)
		if err != nil {
			resultCh <- browserResult{nil, err}
			return
		}

		resultCh <- browserResult{
			browser: playWrightBrowser{
				pw:         res.pw,
				browser:    res.browser,
				userAgent:  res.opt.UserAgent,
				timeout:    *res.opt.Timeout,
				cookieJar:  res.opt.CookieJar,
				ctx:        res.bctx,
				serverAddr: res.opt.ServerAddress,
			},
		}
	}()

	select {
	case <-ctx.Done():
		return nil, ctx.Err()
	case result := <-resultCh:
		return result.browser, result.err
	}
}

func (b playWrightBrowser) updateCookies(_ context.Context, page playwright.Page) error {
	if b.cookieJar != nil {
		cookies, err := page.Context().Cookies(page.URL())
		if err != nil {
			return fmt.Errorf("error getting cookies from browser: %w", err)
		}

		for _, cookie := range cookies {
			// TODO: add support for deleting cookies from the jar which are deleted in the browser
			err = b.cookieJar.Set(playwrightCookieToCookie(cookie))

			if err != nil {
				return fmt.Errorf("error setting cookie in cookie jar: %w", err)
			}
		}
	}

	return nil
}

func (b playWrightBrowser) openPage(_ context.Context, target string, opts OpenPageOptions) (playwright.Page, error) {
	page, err := b.ctx.NewPage()
	if err != nil {
		return nil, err
	}

	pwOpts := playwright.PageGotoOptions{
		WaitUntil: playwright.WaitUntilStateLoad,
	}

	if b.timeout > 0 {
		var ms = float64(b.timeout.Milliseconds())
		pwOpts.Timeout = &ms
	}

	if opts.Referer != "" {
		pwOpts.Referer = playwright.String(opts.Referer)
	}

	resp, err := page.Goto(target, pwOpts)
	if err != nil {
		return nil, err
	}

	slog.Info("opened document", "url", target, "status", resp.Status(), "request", resp.Request())

	if resp.Status() < 200 || resp.Status() >= 300 {
		_ = page.Close()

		if resp.Status() == 404 {
			return nil, ErrPageNotFound
		}
		slog.Info("invalid status code", "status", resp.Status(), "request", resp.Request())
		return nil, fmt.Errorf("%w: %d", ErrInvalidStatusCode, resp.Status())
	}

	return page, nil
}

func (b playWrightBrowser) Open(ctx context.Context, url string, opts OpenPageOptions) (Document, error) {

	page, err := b.openPage(ctx, url, opts)
	if err != nil {
		return nil, err
	}

	err = b.updateCookies(ctx, page)
	if err != nil {
		return nil, err
	}

	return newDocument(b.pw, b.browser, page)
}

func (b playWrightBrowser) Close() error {
	return errors.Join(
		b.browser.Close(),
		b.ctx.Close(),
		b.pw.Stop(),
	)
}


func Screenshot(ctx context.Context, target string, timeout time.Duration) ([]byte, error) {
	browser, err := NewBrowser(ctx, BrowserOptions{
		Timeout: &timeout,
	})

	if err != nil {
		return nil, fmt.Errorf("error creating browser: %w", err)
	}

	defer DeferClose(browser)

	doc, err := browser.Open(ctx, target, OpenPageOptions{})
	if err != nil {
		return nil, fmt.Errorf("error opening page: %w", err)
	}

	defer DeferClose(doc)

	return doc.Screenshot()
}