Skip to content

Instantly share code, notes, and snippets.

@sergiolopes
Created September 12, 2011 00:45
Show Gist options
  • Star 3 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save sergiolopes/1210374 to your computer and use it in GitHub Desktop.
Save sergiolopes/1210374 to your computer and use it in GitHub Desktop.
Bastidores das análises dos Sites dos participantes do QCon SP
Algumas pessoas me perguntaram como fiz a análise dos Sites do pessoal que foi no QCon. Sim, eu analisei
o Site de todo mundo, um por um, e fiz as médias. Mas obviamente, nada manual, então lá vai como fiz:
1) Peguei o e-mail de todo mundo cadastrado e extraí o domínio. Tirei os e-mails de webmails genéricos
(gmail, hotmail etc) e separei os domínios das empresas. Salvei tudo num urls.txt (no total deram umas
100 URLs únicas).
2) Submeti todos os domínios (prefixados por http://www) pro WebPageTest.org. Para isso, fiz um programinha
Java usando Selenium2 (código abaixo no gist)
3) O programa anterior cospe os test_ids gerados pelo WebPageTest para cada teste. Esperei uns minutos e
baixei os resultados (em .csv e .har) usando outro programinha Java (código abaixo no gist).
4) Peguei os CSV e joguei num Google Spreadsheets pra ordenar, tirar médias de algumas métricas etc.
PS. O WebPageTest tem uma API REST mas você precisa registrar um key o que não é muito fácil (ou rodar
sua instância própria no EC2). Usando o Selenium consegui submeter os testes sem problemas usando a
interface normal do Site mesmo :)
PS2: Nao vou divulgar os endereços individuais nem estatisticas de sites especificos. Mas se alguem
quiser saber mais algum numero de media, so perguntar
-----
Slides da apresentação com resultado dos estudos:
www.slideshare.net/caelumdev/qcon-2011-por-uma-web-mais-rpida-tcnicas-de-otimizao-de-sites
// "scriptizao" pra baixar os resultados depois de executados
public class DownloadResults {
public static void main(String[] args) throws FileNotFoundException {
Scanner results = new Scanner(new File("test-ids.txt"));
while (results.hasNextLine()) {
String[] result = results.nextLine().split("\\s+");
String testId = result[0];
String domain = result[1];
System.out.print("Downloading results for " + domain);
String pageDataUri = "http://www.webpagetest.org/result/" + testId + "/" + testId + "_www." + domain + "_page_data.csv";
String pageDataOutput = "results/page_data/" + domain + ".csv";
try {
download(pageDataUri, pageDataOutput);
} catch (Exception e) {
e.printStackTrace();
}
String requestsUri = "http://www.webpagetest.org/result/" + testId + "/" + testId + "_www." + domain + "_requests.csv";
String requestsOutput = "results/requests/" + domain + ".csv";
try {
download(requestsUri, requestsOutput);
} catch (Exception e) {
e.printStackTrace();
}
String harUri = "http://www.webpagetest.org/export.php?test=" + testId;
String harOutput = "results/har/" + domain + ".json";
try {
download(harUri, harOutput);
} catch (Exception e) {
e.printStackTrace();
}
System.out.println(" [COMPLETED]");
}
results.close();
}
private static void download(String uri, String output)
throws MalformedURLException, IOException, FileNotFoundException {
URL url = new URL(uri);
ReadableByteChannel rbc = Channels.newChannel(url.openStream());
FileOutputStream fos = new FileOutputStream(output);
fos.getChannel().transferFrom(rbc, 0, 1 << 24);
fos.close();
rbc.close();
}
}
// "scriptizão" feio pra submeter os testes no WebPageTest.org
// pra rodar, tenha os JARs do Selenium2 no classpath
public class PostTests {
static FirefoxDriver driver = new FirefoxDriver();
public static void main(String[] args) throws FileNotFoundException {
PrintWriter tests = new PrintWriter("test-ids.txt");
Scanner domains = new Scanner(new File("urls.txt"));
while (domains.hasNextLine()) {
try {
String domain = domains.nextLine();
String uri = "http://www." + domain + "/";
System.out.print("Testing: " + uri);
String testId = submitTestTo(uri);
System.out.println(" [COMPLETED]");
tests.println(testId + " " + domain);
tests.flush();
} catch (Exception e) {
e.printStackTrace();
continue;
}
}
domains.close();
tests.close();
driver.quit();
}
private static String submitTestTo(String uri) {
driver.navigate().to("http://www.webpagetest.org/");
WebElement uriField = driver.findElement(By.name("url"));
uriField.clear();
uriField.sendKeys(uri);
driver.findElement(By.name("submit")).click();
(new WebDriverWait(driver, 15)).until(new ExpectedCondition<Boolean>() {
public Boolean apply(WebDriver d) {
return d.getCurrentUrl().startsWith("http://www.webpagetest.org/result/");
}
});
String testUrl = driver.getCurrentUrl();
String testId = testUrl.split("/")[4];
return testId;
}
}
@zeroum
Copy link

zeroum commented Sep 12, 2011

Fiz um teste naquelas lojas com home de 8 MB com o Chrome (Incognito mode para não cachear), e elas deram pouco mais de 2 MB. Talvez tenham trocado o site deles de lá pra cá, ou seja um bug no Web Page Test.

@ericsaboia
Copy link

Muito legal! cade o post com as referências que você falou que iria publicar?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment