Add monkeypatching

This commit is contained in:
Adrian Vollmer 2022-09-25 16:53:14 +02:00
parent a34cfa6027
commit 2489e8ab31
4 changed files with 133 additions and 23 deletions

View File

@ -38,6 +38,7 @@ def embed_assets(index_file):
'inject.js',
'init.css',
'init.html',
'monkeypatch.js',
'pako.min.js',
]:
path = os.path.join(SCRIPT_PATH, filename)
@ -48,7 +49,12 @@ def embed_assets(index_file):
new_base_name = 'SELF_CONTAINED_' + base_name
result_file = os.path.join(base_dir, new_base_name)
file_tree = load_filetree(base_dir, init_files['inject.js'], exclude_pattern=new_base_name)
file_tree = load_filetree(
base_dir,
before=init_files['monkeypatch.js'],
after=init_files['inject.js'],
exclude_pattern=new_base_name,
)
file_tree = json.dumps(file_tree)
logger.debug('total asset size: %d' % len(file_tree))
file_tree = deflate(file_tree)
@ -86,7 +92,7 @@ def embed_assets(index_file):
return result_file
def pack_file(filename, js):
def pack_file(filename, before, after):
_, ext = os.path.splitext(filename)
ext = ext.lower()[1:]
data = open(filename, 'rb').read()
@ -101,7 +107,12 @@ def pack_file(filename, js):
data = base64.b64encode(data)
elif ext in ['html', 'htm']:
data = embed_html_resources(data, os.path.dirname(filename), js).encode()
data = embed_html_resources(
data,
os.path.dirname(filename),
before,
after,
).encode()
if not isinstance(data, str):
try:
@ -120,16 +131,22 @@ def deflate(data):
return data
def embed_html_resources(html, base_dir, js):
def embed_html_resources(html, base_dir, before, after):
"""Embed fonts in preload links to avoid jumps when loading"""
# This cannot be done in JavaScript, it would be too late
import bs4
soup = bs4.BeautifulSoup(html, 'lxml')
script = soup.new_tag("script")
script.string = js
soup.find('body').append(script)
if before:
script = soup.new_tag("script")
script.string = before
soup.find('body').insert(0, script)
if after:
script = soup.new_tag("script")
script.string = after
soup.find('body').append(script)
return str(soup)
@ -202,7 +219,7 @@ def embed_css_resources(css, filename):
return css
def load_filetree(base_dir, js, exclude_pattern=None):
def load_filetree(base_dir, before=None, after=None, exclude_pattern=None):
"""Load entire directory in a dict"""
result = {}
@ -214,7 +231,8 @@ def load_filetree(base_dir, js, exclude_pattern=None):
key = path.relative_to(base_dir).as_posix()
result[key] = pack_file(
path.as_posix(),
js,
before,
after,
)
logger.debug('Packed file %s [%d]' % (key, len(result[key])))

View File

@ -32,13 +32,41 @@ var createIframe = function() {
return iframe;
}
// var load_blob = (function () {
// var a = document.createElement("a");
// document.body.appendChild(a);
// a.style = "display: none";
// return function (data, get_params, anchor) {
// var blob = new Blob([data], {type: "text/html"}),
// url = window.URL.createObjectURL(blob);
// if (get_params) {
// url += "?" + get_params; // This is considered a security issue
// }
// if (anchor) {
// url += "#" + anchor;
// }
// a.href = url;
// a.target = 'main';
// a.click();
// window.URL.revokeObjectURL(url);
// };
// }());
//
// var load_virtual_page = (function (path, get_params, anchor) {
// const data = window.data.file_tree[path];
// var iframe = createIframe();
// load_blob(data, get_params, anchor);
// window.data.current_path = path;
// });
var load_virtual_page = (function (path, get_params, anchor) {
const data = window.data.file_tree[path];
var iframe = createIframe();
if (get_params) { iframe.src = path + '?' + get_params; }
else { iframe.src = path; }
window.data.get_parameters = get_params;
iframe.contentDocument.write(data);
if (anchor) { iframe.contentDocument.location.hash = anchor; }
if (anchor) {
iframe.contentDocument.location.hash = anchor;
}
window.data.current_path = path;
});

View File

@ -51,8 +51,17 @@ var split_url = function(url) {
var virtual_click = function(evnt) {
// Handle GET parameters and anchors
console.log("Virtual click", evnt);
var a = evnt.currentTarget;
let [path, get_parameters, anchor] = split_url(a.getAttribute('href'));
var el = evnt.currentTarget;
var name = el.tagName.toLowerCase();
if (name == 'a') {
var [path, get_parameters, anchor] = split_url(el.getAttribute('href'));
} else if (name == 'form') {
var [path, get_parameters, anchor] = split_url(el.getAttribute('action'));
const formData = new FormData(el);
get_parameters = new URLSearchParams(formData).toString();
} else {
console.error("Invalid element", el);
}
path = normalize_path(path);
window.parent.postMessage({
@ -79,14 +88,8 @@ var fix_links = function(origin) {
var fix_forms = function(origin) {
Array.from(document.querySelectorAll("form")).forEach( form => {
var href = form.getAttribute('action');
if (is_virtual(href)) {
// TODO test this
// let [path, get_parameters, anchor] = split_url(href);
// path = normalize_path(path);
// var new_href = to_blob(retrieve_file(path), 'text/html');
// if (get_parameters) { new_href += '?' + get_parameters; }
// if (anchor) { new_href += '?' + anchor; }
// form.action = new_href;
if (is_virtual(href) && form.getAttribute('method').toLowerCase() == 'get') {
form.addEventListener('submit', virtual_click);
}
});
};
@ -164,11 +167,17 @@ window.addEventListener("message", (evnt) => {
console.log("Received data from parent", window.data);
// dynamically fix elements on this page
try {
embed_js(); // This might change the DOM, so do this first
embed_css();
fix_links();
fix_forms();
embed_img();
embed_js();
// Trigger DOMContentLoaded again, some scripts that have just
// been executed expect it.
window.document.dispatchEvent(new Event("DOMContentLoaded", {
bubbles: true,
cancelable: true
}));
} finally {
window.parent.postMessage({
action: "show_iframe",
@ -178,6 +187,7 @@ window.addEventListener("message", (evnt) => {
}
}, false);
// Set parent window title
window.parent.postMessage({
action: "set_title",

54
src/monkeypatch.js Normal file
View File

@ -0,0 +1,54 @@
/*
* Monkeypatch URLSearchParams
*
* Sphinx documents that use `searchtool.js` rely on passing information via
* GET parameters (aka search parameters). Unfortunately, this doesn't work
* in our approach due to the same origin policy, so we have to get ...
* creative.
*
* Here, we patch the `URLSearchParams` class so it returns the information
* stored in `window.data.get_parameters`.
*
*/
const originalGet = URLSearchParams.prototype.get;
var myGet = function (arg) {
const originalResult = originalGet.apply(this, [arg]);
// If searchtools.js of sphinx is used
if (
(arg == "q" || arg == "highlight") &&
window.DOCUMENTATION_OPTIONS &&
window.Scorer &&
window.data.get_parameters &&
(! originalResult)
) {
const params = new URLSearchParams('?' + window.data.get_parameters);
const result = params.get("q");
return result;
} else {
return originalResult;
}
};
URLSearchParams.prototype.get = myGet;
/*
* Monkeypatch fetch
*/
const { fetch: originalFetch } = window;
window.fetch = async (...args) => {
let [resource, config ] = args;
var path = normalize_path(resource);
var response;
if (is_virtual(path)) {
var data = retrieve_file(path);
response = new Response(data);
} else {
response = await originalFetch(resource, config);
}
return response;
};