go-extractor/document.go

package extractor

import (
	"fmt"
	"io"
	"log/slog"
	"time"

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

type Document interface {
	io.Closer
	Node

	URL() string
	Refresh() error
	Content() (string, error)

	WaitForNetworkIdle(timeout *time.Duration) error
}

type document struct {
	node
	pw      *playwright.Playwright
	browser playwright.Browser
	page    playwright.Page
	root    playwright.ElementHandle
	locator playwright.Locator
}

func newDocument(pw *playwright.Playwright, browser playwright.Browser, page playwright.Page) (Document, error) {
	root, err := page.QuerySelector("html")

	if err != nil {
		return nil, err
	}

	root2 := page.Locator("html")

	res := &document{
		node: node{
			locator: root2,
		},
		pw:      pw,
		browser: browser,
		page:    page,
		root:    root,
	}

	slog.Info("new document", "url", page.URL(), "root", root, "locator", root2)

	return res, nil
}
func (d *document) Close() error {
	return d.page.Close()
}

func (d *document) URL() string {
	return d.page.URL()
}

func (d *document) Content() (string, error) {
	return d.page.Content()
}

func (d *document) Refresh() error {
	resp, err := d.page.Reload()
	if err != nil {
		return fmt.Errorf("failed to reload page: %w", err)
	}

	if resp.Status() != 200 {
		return fmt.Errorf("invalid status code: %d", resp.Status())
	}

	return nil
}

func (d *document) WaitForNetworkIdle(timeout *time.Duration) error {

	var f *float64 = nil
	if timeout == nil {
		t := 30 * time.Second
		timeout = &t
	}

	if timeout != nil {
		ms := float64(timeout.Milliseconds())
		f = &ms
	}

	err := d.page.WaitForLoadState(playwright.PageWaitForLoadStateOptions{
		State:   playwright.LoadStateNetworkidle,
		Timeout: f,
	})
	return err
}