Switch from own docker image to standard selenium images

Since the Selenium images are available for download, and downloading them
is usually faster than building them from scratch, this makes taking
screenshots easier.  Furthermore, since the Selenium image is not specific
to KaTeX, it could as well be used for other purposes, thus saving space
since a single image can be used in multiple projects.

This change also deals with the non-determinism in the Lap screenshot:
We detect the one known (and accepted) alternate rendering and change the
output file name to Lap_alt in this case.  So either Lap or Lap_alt gets
saved to, and if the image is different from both, then one of these files
will show a modification.  On the other hand, if it is equal to either of
these, then the matching one will get overwritten, showing no change.
This commit is contained in:
Martin von Gagern 2015-06-22 19:30:26 +02:00
parent b1d5311898
commit 5d155c75db
8 changed files with 326 additions and 141 deletions

View File

@ -1,14 +0,0 @@
FROM ubuntu:14.04
MAINTAINER xymostech <xymostech@gmail.com>
RUN apt-get -qq update
RUN apt-get -qqy install default-jre=2:1.7-51 firefox=28.0+build2-0ubuntu2 xvfb=2:1.15.1-0ubuntu2 wget=1.15-1ubuntu1 python=2.7.5-5ubuntu3 python-pip=1.5.4-1 nodejs=0.10.25~dfsg2-2ubuntu1 || true
RUN wget http://selenium-release.storage.googleapis.com/2.43/selenium-server-standalone-2.43.0.jar
RUN ln -s /usr/bin/nodejs /usr/bin/node
RUN pip install selenium pypng
ENV DISPLAY :1
CMD /bin/bash ~/run.sh
RUN echo "java -jar /selenium-server-standalone-2.43.0.jar > /dev/null &" >> ~/run.sh
RUN echo "Xvfb :1 2> /dev/null &" >> ~/run.sh
RUN echo "make -C /KaTeX serve > /dev/null &" >> ~/run.sh
RUN echo "sleep 2" >> ~/run.sh
RUN echo "/KaTeX/dockers/Screenshotter/screenshotter.py /KaTeX/test/screenshotter/ss_data.json" >> ~/run.sh

View File

@ -1,29 +1,65 @@
### How to generate screenshotter images
----------------------------------------
# How to generate screenshotter images
## Automatic generation of screen shots
Now you too can generate screenshots from your own computer, and (hopefully)
have them look mostly the same as the current ones! To start, make a docker
image from the included Dockerfile using a command like
have them look mostly the same as the current ones! Make sure you have docker
installed and running. Also make sure that the development server is running,
or start it by running
docker build --tag=ss .
node server.js
from within this directory (note you need to have docker installed and running
for this to work). This will build a docker image with the `ss` tag, which you
can then use to run dockers based on it.
in the top level directory of the source tree. If all you want is (re)create
all the snapshots for all the browsers, then you can do so by running the
`screenshotter.sh` script:
This Dockerfile is set up such that it will run everything and generate all the
screenshots when the docker is run, so no interactive input is required. All
that you need to do is mount the KaTeX directory you want to test into the
`/KaTeX` directory in the docker, and run the `ss` docker, like so:
dockers/Screenshotter/screenshotter.sh
docker run --volume=/your/KaTeX/:/KaTeX ss
It will fetch all required selenium docker images, and use them to
take screenshots.
The `--volume=/your/KaTeX:/KaTeX` switch mounts your KaTeX directory into the
docker. Note this is a read-write mounting, so the new screenshots will be
directly placed into your KaTeX directory.
## Manual generation
Since this docker is very self-contained, there should be no need to do
interactive management of the docker, but if you feel the need, you can read the
General Docker Help section of the MathJaxFonts docker readme.
If you are creating screenshots on a regular basis, you can keep the
docker containers with the selenium setups running. Essentially you
are encouraged to reproduce the steps from `screenshotter.sh`
manually. Example run for Firefox:
That's it!
container=$(docker run -d -P selenium/standalone-firefox:2.46.0)
node dockers/Screenshotter/screenshotter.js -b firefox -c ${container}
# possibly repeat the above command as often as you need, then eventually
docker stop ${container}
docker rm ${container}
## Use without docker
It is possible to run `screenshotter.js` without the use of Docker:
npm install selenium-webdriver
node dockers/Screenshotter/screenshotter.js
This will generate screenshots using the Firefox installed on your system.
Browsers other than Firefox can be targeted using the `--browser` option.
For a complete list of options pass `--help` as an argument to
`screenshotter.js`. Using these it should be possible to have the script
connect to almost any Selenium web driver you might have access to.
Note that screenshots taken without Docker are very likely to disagree
from the ones stored in the repository, due to different versions of
various software components being used. The screenshots taken in this
fashion are well suited for visual inspection, but for exact binary
comparisons it would be neccessary to carefully set up the environment
to match the one used by the Docker approach.
## Choosing the list of test cases
Both `screenshotter.js` and `screenshotter.sh` will accept
an `--include` option (short `-i`) which can be used to specify
a list of test cases to be processed, as a comma separated list.
Conversely, the `--exclude` option (short `-x`) can be used
to specify a list of cases which are not being processed.
Examples:
node dockers/Screenshotter/screenshotter.js -i Sqrt,SqrtRoot
dockers/Screenshotter/screenshotter.sh --exclude=GreekLetters

View File

@ -0,0 +1,237 @@
"use strict";
var childProcess = require("child_process");
var fs = require("fs");
var path = require("path");
var net = require("net");
var selenium = require("selenium-webdriver");
var data = require("../../test/screenshotter/ss_data.json");
var dstDir = path.normalize(
path.join(__dirname, "..", "..", "test", "screenshotter", "images"));
//////////////////////////////////////////////////////////////////////
// Process command line arguments
var opts = require("nomnom")
.option("browser", {
abbr: "b",
"default": "firefox",
help: "Name of the browser to use"
})
.option("container", {
abbr: "c",
type: "string",
help: "Name or ID of a running docker container to contact"
})
.option("seleniumURL", {
full: "selenium-url",
help: "Full URL of the Selenium web driver"
})
.option("seleniumIP", {
full: "selenium-ip",
help: "IP address of the Selenium web driver"
})
.option("seleniumPort", {
full: "selenium-port",
"default": 4444,
help: "Port number of the Selenium web driver"
})
.option("katexURL", {
full: "katex-url",
help: "Full URL of the KaTeX development server"
})
.option("katexIP", {
full: "katex-ip",
"default": "localhost",
help: "Full URL of the KaTeX development server"
})
.option("katexPort", {
full: "katex-port",
"default": 7936,
help: "Port number of the KaTeX development server"
})
.option("include", {
abbr: "i",
help: "Comma-separated list of test cases to process"
})
.option("exclude", {
abbr: "x",
help: "Comma-separated list of test cases to exclude"
})
.parse();
var listOfCases;
if (opts.include) {
listOfCases = opts.include.split(",");
} else {
listOfCases = Object.keys(data);
}
if (opts.exclude) {
var exclude = opts.exclude.split(",");
listOfCases = listOfCases.filter(function(key) {
return exclude.indexOf(key) === -1;
});
}
var seleniumURL = opts.seleniumURL;
var katexURL = opts.katexURL;
var seleniumIP = opts.seleniumIP;
var seleniumPort = opts.seleniumPort;
var katexIP = opts.katexIP;
//////////////////////////////////////////////////////////////////////
// Work out connection to selenium docker container
function check(err) {
if (!err) {
return;
}
console.error(err);
console.error(err.stack);
process.exit(1);
}
function dockerCmd() {
var args = Array.prototype.slice.call(arguments);
return childProcess.execFileSync(
"docker", args, { encoding: "utf-8" }).replace(/\n$/, "");
}
if (!seleniumURL && opts.container) {
try {
// When using boot2docker, seleniumIP and katexIP are distinct.
seleniumIP = childProcess.execFileSync(
"boot2docker", ["ip"], { encoding: "utf-8" }).replace(/\n$/, "");
var config = childProcess.execFileSync(
"boot2docker", ["config"], { encoding: "utf-8" });
config = (/^HostIP = "(.*)"$/m).exec(config);
if (!config) {
console.error("Failed to find HostIP");
process.exit(2);
}
katexIP = config[1];
} catch(e) {
seleniumIP = katexIP = dockerCmd(
"inspect", "-f", "{{.NetworkSettings.Gateway}}", opts.container);
}
seleniumPort = dockerCmd("port", opts.container, seleniumPort);
seleniumPort = seleniumPort.replace(/^.*:/, "");
}
if (!seleniumURL && seleniumIP) {
seleniumURL = "http://" + seleniumIP + ":" + seleniumPort + "/wd/hub";
}
if (seleniumURL) {
console.log("Selenium driver at " + seleniumURL);
} else {
console.log("Selenium driver in local session");
}
if (!katexURL) {
katexURL = "http://" + katexIP + ":" + opts.katexPort + "/";
}
var toStrip = "http://localhost:7936/"; // remove this from testcase URLs
//////////////////////////////////////////////////////////////////////
// Wait for container to become ready
var attempts = 0;
process.nextTick(seleniumIP ? tryConnect : buildDriver);
function tryConnect() {
var sock = net.connect({
host: seleniumIP,
port: +seleniumPort
});
sock.on("connect", function() {
sock.end();
attempts = 0;
setTimeout(buildDriver, 0);
}).on("error", function() {
if (++attempts > 50) {
throw new Error("Failed to connect selenium server.");
}
setTimeout(tryConnect, 200);
});
}
//////////////////////////////////////////////////////////////////////
// Build the web driver
var driver;
function buildDriver() {
var builder = new selenium.Builder().forBrowser(opts.browser);
if (seleniumURL) {
builder.usingServer(seleniumURL);
}
driver = builder.build();
setSize(targetW, targetH);
}
//////////////////////////////////////////////////////////////////////
// Set the screen size
var targetW = 1024, targetH = 768;
function setSize(reqW, reqH) {
return driver.manage().window().setSize(reqW, reqH).then(function() {
return driver.takeScreenshot();
}).then(function(img) {
img = imageDimensions(img);
var actualW = img.width;
var actualH = img.height;
if (actualW === targetW && actualH === targetH) {
process.nextTick(takeScreenshots);
return;
}
if (++attempts > 5) {
throw new Error("Failed to set window size correctly.");
}
return setSize(targetW + reqW - actualW, targetH + reqH - actualH);
}, check);
}
function imageDimensions(img) {
var buf = new Buffer(img, "base64");
return {
buf: buf,
width: buf.readUInt32BE(16),
height: buf.readUInt32BE(20)
};
}
//////////////////////////////////////////////////////////////////////
// Take the screenshots
function takeScreenshots() {
listOfCases.forEach(takeScreenshot);
}
function takeScreenshot(key) {
var url = data[key];
if (!url) {
console.error("Test case " + key + " not known!");
return;
}
url = katexURL + url.substr(toStrip.length);
driver.get(url);
driver.takeScreenshot().then(function haveScreenshot(img) {
img = imageDimensions(img);
if (img.width !== targetW || img.height !== targetH) {
throw new Error("Excpected " + targetW + " x " + targetH +
", got " + img.width + "x" + img.height);
}
if (key === "Lap" && opts.browser === "firefox" &&
img.buf[0x32] === 0xf8) {
/* There is some strange non-determinism with this case,
* causing slight vertical shifts. The first difference
* is at offset 0x32, where one file has byte 0xf8 and
* the other has something else. By using a different
* output file name for one of these cases, we accept both.
*/
key += "_alt";
}
var file = path.join(dstDir, key + "-" + opts.browser + ".png");
fs.writeFile(file, img.buf, check);
console.log(key);
}, check);
}

View File

@ -1,107 +0,0 @@
#!/usr/bin/env python2
import argparse
import json
import os
import png
import StringIO
import sys
from selenium import webdriver
def get_png_size(png_data):
w, h, _, _ = png.Reader(file=StringIO.StringIO(png_data)).read()
return (w, h)
def set_driver_size(driver, width, height):
"""Correctly sets the size of the driver window so screenshots end up the
provided size"""
driver.set_window_size(width, height)
screenshot_size = get_png_size(driver.get_screenshot_as_png())
attempts = 0
while (width, height) != screenshot_size:
attempts += 1
if attempts > 5:
print "Tried 5 times to size screen correctly, bailing out"
exit(1)
ss_width, ss_height = screenshot_size
driver.set_window_size(
width + (width - ss_width),
height + (height - ss_height))
screenshot_size = get_png_size(driver.get_screenshot_as_png())
def main():
parser = argparse.ArgumentParser(
description='Take screenshots of webpages', add_help=False)
parser.add_argument('file', metavar='file.json')
parser.add_argument('-t', '--tests', metavar='test', nargs='*')
parser.add_argument('-w', '--width', metavar='width', default=1024,
type=int)
parser.add_argument('-h', '--height', metavar='height', default=768,
type=int)
parser.add_argument('-b', '--browser', metavar='browser',
choices=['firefox'], default='firefox')
args = parser.parse_args()
data = None
with open(args.file) as f:
try:
data = json.load(f)
except ValueError:
print "Invalid json in input file:", args.file
exit(1)
tests = []
if args.tests is None:
tests = data.keys()
else:
data_tests = data.keys()
for test in args.tests:
if test not in data_tests:
print "Unknown test:", test
exit(1)
tests = args.tests
print "Starting up"
sys.stdout.flush()
driver = None
if args.browser == 'firefox':
driver = webdriver.Firefox()
else:
print "Unknown browser:", args.browser
exit(1)
set_driver_size(driver, args.width, args.height)
data_dir = os.path.join(
os.path.dirname(os.path.realpath(args.file)), "images")
try:
os.mkdir(data_dir)
except OSError:
pass
for test, url in data.iteritems():
if test in tests:
filename = os.path.join(
data_dir, '%s-%s.png' % (test, args.browser))
print "Running:", test
sys.stdout.flush()
driver.get(url)
driver.get_screenshot_as_file(filename)
print "Done"
if __name__ == '__main__':
main()

View File

@ -0,0 +1,28 @@
#!/bin/bash
# This script does a one-shot creation of screenshots, creating needed
# docker containers and removing them afterwards. During development,
# it might be desirable to avoid the overhead for starting and
# stopping the containers. Developers are encouraged to manage
# suitable containers themselves, calling the screenshotter.js script
# directly.
status=0
for browserTag in firefox:2.46.0; do
browser=${browserTag%:*}
image=selenium/standalone-${browserTag}
echo "Starting container for ${image}"
container=$(docker run -d -P ${image})
[[ ${container} ]] || continue
echo "Container ${container:0:12} started, creating screenshots..."
if node "$(dirname "$0")"/screenshotter.js \
--browser="${browser}" --container="${container}" "$@"; then
res=Done
else
res=Failed
status=1
fi
echo "${res} taking screenshots, stopping and removing ${container:0:12}"
docker stop ${container} >/dev/null && docker rm ${container} >/dev/null
done
exit ${status}

View File

@ -20,6 +20,8 @@
"jasmine-node": "2.0.0-beta4",
"jshint": "^2.5.6",
"less": "~1.7.5",
"selenium-webdriver": "^2.46.1",
"nomnom": "^1.8.1",
"uglify-js": "~2.4.15"
},
"bin": "cli.js",

Binary file not shown.

After

Width:  |  Height:  |  Size: 12 KiB

View File

@ -8,6 +8,9 @@
#math, #pre, #post {
font-size: 4em;
}
body {
font-family: "DejaVu Serif",serif;
}
</style>
</head>
<body>