Files
parcer/maxun-core/src/interpret.ts

616 lines
22 KiB
TypeScript
Raw Normal View History

/* eslint-disable no-await-in-loop, no-restricted-syntax */
2024-08-03 00:28:54 +05:30
import { Page, PageScreenshotOptions } from 'playwright';
2024-08-21 05:25:54 +05:30
import { PlaywrightBlocker } from '@cliqz/adblocker-playwright';
2024-08-21 05:06:34 +05:30
import fetch from 'cross-fetch';
import path from 'path';
import { EventEmitter } from 'events';
import {
Where, What, PageState, Workflow, WorkflowFile,
ParamType, SelectorArray, CustomFunctions,
} from './types/workflow';
import { operators, meta } from './types/logic';
import { arrayToObject } from './utils/utils';
import Concurrency from './utils/concurrency';
import Preprocessor from './preprocessor';
import log, { Level } from './utils/logger';
/**
* Defines optional intepreter options (passed in constructor)
*/
interface InterpreterOptions {
maxRepeats: number;
maxConcurrency: number;
serializableCallback: (output: any) => (void | Promise<void>);
binaryCallback: (output: any, mimeType: string) => (void | Promise<void>);
debug: boolean;
debugChannel: Partial<{
activeId: Function,
debugMessage: Function,
}>
}
/**
* Class for running the Smart Workflows.
*/
export default class Interpreter extends EventEmitter {
private workflow: Workflow;
private initializedWorkflow: Workflow | null;
private options: InterpreterOptions;
private concurrency: Concurrency;
private stopper: Function | null = null;
private log: typeof log;
2024-08-21 05:06:34 +05:30
private blocker: PlaywrightBlocker | null = null;
constructor(workflow: WorkflowFile, options?: Partial<InterpreterOptions>) {
super();
this.workflow = workflow.workflow;
this.initializedWorkflow = null;
this.options = {
maxRepeats: 5,
maxConcurrency: 5,
serializableCallback: (data) => { log(JSON.stringify(data), Level.WARN); },
binaryCallback: () => { log('Received binary data, thrashing them.', Level.WARN); },
debug: false,
debugChannel: {},
...options,
};
this.concurrency = new Concurrency(this.options.maxConcurrency);
this.log = (...args) => log(...args);
const error = Preprocessor.validateWorkflow(workflow);
if (error) {
throw (error);
}
if (this.options.debugChannel?.debugMessage) {
const oldLog = this.log;
// @ts-ignore
this.log = (...args: Parameters<typeof oldLog>) => {
if (args[1] !== Level.LOG) {
this.options.debugChannel.debugMessage!(typeof args[0] === 'string' ? args[0] : args[0].message);
}
oldLog(...args);
};
}
2024-08-21 05:10:56 +05:30
2024-08-21 05:11:28 +05:30
PlaywrightBlocker.fromPrebuiltAdsAndTracking(fetch).then(blocker => {
2024-08-21 05:10:56 +05:30
this.blocker = blocker;
2024-08-21 05:11:28 +05:30
}).catch(err => {
2024-08-21 05:10:56 +05:30
this.log(`Failed to initialize ad-blocker:`, Level.ERROR);
})
}
2024-08-21 05:14:58 +05:30
private async applyAdBlocker(page: Page): Promise<void> {
if (this.blocker) {
await this.blocker.enableBlockingInPage(page);
}
}
2024-08-21 05:24:41 +05:30
private async disableAdBlocker(page: Page): Promise<void> {
if (this.blocker) {
2024-08-21 05:24:52 +05:30
await this.blocker.disableBlockingInPage(page);
2024-08-21 05:24:41 +05:30
}
2024-08-21 05:24:52 +05:30
}
2024-08-21 05:24:41 +05:30
/**
* Returns the context object from given Page and the current workflow.\
* \
* `workflow` is used for selector extraction - function searches for used selectors to
* look for later in the page's context.
* @param page Playwright Page object
* @param workflow Current **initialized** workflow (array of where-what pairs).
* @returns {PageState} State of the current page.
*/
private async getState(page: Page, workflow: Workflow): Promise<PageState> {
2024-10-24 18:47:18 +05:30
await page.setViewportSize({ width: 900, height: 400 });
/**
* All the selectors present in the current Workflow
*/
const selectors = Preprocessor.extractSelectors(workflow);
/**
* Determines whether the element targetted by the selector is [actionable](https://playwright.dev/docs/actionability).
* @param selector Selector to be queried
* @returns True if the targetted element is actionable, false otherwise.
*/
const actionable = async (selector: string): Promise<boolean> => {
try {
const proms = [
page.isEnabled(selector, { timeout: 500 }),
page.isVisible(selector, { timeout: 500 }),
];
return await Promise.all(proms).then((bools) => bools.every((x) => x));
} catch (e) {
// log(<Error>e, Level.ERROR);
return false;
}
};
/**
* Object of selectors present in the current page.
*/
const presentSelectors: SelectorArray = await Promise.all(
selectors.map(async (selector) => {
if (await actionable(selector)) {
return [selector];
}
return [];
}),
).then((x) => x.flat());
return {
url: page.url(),
cookies: (await page.context().cookies([page.url()]))
.reduce((p, cookie) => (
{
...p,
[cookie.name]: cookie.value,
}), {}),
selectors: presentSelectors,
};
}
/**
* Tests if the given action is applicable with the given context.
* @param where Tested *where* condition
* @param context Current browser context.
* @returns True if `where` is applicable in the given context, false otherwise
*/
private applicable(where: Where, context: PageState, usedActions: string[] = []): boolean {
/**
* Given two arbitrary objects, determines whether `subset` is a subset of `superset`.\
* \
* For every key in `subset`, there must be a corresponding key with equal scalar
* value in `superset`, or `inclusive(subset[key], superset[key])` must hold.
* @param subset Arbitrary non-cyclic JS object (where clause)
* @param superset Arbitrary non-cyclic JS object (browser context)
* @returns `true` if `subset <= superset`, `false` otherwise.
*/
const inclusive = (subset: Record<string, unknown>, superset: Record<string, unknown>)
: boolean => (
Object.entries(subset).every(
([key, value]) => {
/**
* Arrays are compared without order (are transformed into objects before comparison).
*/
const parsedValue = Array.isArray(value) ? arrayToObject(value) : value;
const parsedSuperset: Record<string, unknown> = {};
parsedSuperset[key] = Array.isArray(superset[key])
? arrayToObject(<any>superset[key])
: superset[key];
// Every `subset` key must exist in the `superset` and
// have the same value (strict equality), or subset[key] <= superset[key]
return parsedSuperset[key]
&& (
(parsedSuperset[key] === parsedValue)
|| ((parsedValue).constructor.name === 'RegExp' && (<RegExp>parsedValue).test(<string>parsedSuperset[key]))
|| (
(parsedValue).constructor.name !== 'RegExp'
&& typeof parsedValue === 'object' && inclusive(<typeof subset>parsedValue, <typeof superset>parsedSuperset[key])
)
);
},
)
);
// Every value in the "where" object should be compliant to the current state.
return Object.entries(where).every(
([key, value]) => {
if (operators.includes(<any>key)) {
const array = Array.isArray(value)
? value as Where[]
: Object.entries(value).map((a) => Object.fromEntries([a]));
// every condition is treated as a single context
2024-09-21 22:20:04 +05:30
switch (key as keyof typeof operators) {
case '$and':
return array?.every((x) => this.applicable(x, context));
case '$or':
return array?.some((x) => this.applicable(x, context));
case '$not':
return !this.applicable(<Where>value, context); // $not should be a unary operator
default:
throw new Error('Undefined logic operator.');
}
} else if (meta.includes(<any>key)) {
const testRegexString = (x: string) => {
if (typeof value === 'string') {
return x === value;
}
return (<RegExp><unknown>value).test(x);
};
switch (key as keyof typeof meta) {
case '$before':
return !usedActions.find(testRegexString);
case '$after':
return !!usedActions.find(testRegexString);
default:
throw new Error('Undefined meta operator.');
}
} else {
// Current key is a base condition (url, cookies, selectors)
return inclusive({ [key]: value }, context);
}
},
);
}
/**
* Given a Playwright's page object and a "declarative" list of actions, this function
* calls all mentioned functions on the Page object.\
* \
* Manipulates the iterator indexes (experimental feature, likely to be removed in
2024-09-21 22:20:04 +05:30
* the following versions of maxun-core)
* @param page Playwright Page object
* @param steps Array of actions.
*/
private async carryOutSteps(page: Page, steps: What[]): Promise<void> {
/**
* Defines overloaded (or added) methods/actions usable in the workflow.
* If a method overloads any existing method of the Page class, it accepts the same set
* of parameters *(but can override some!)*\
* \
* Also, following piece of code defines functions to be run in the browser's context.
* Beware of false linter errors - here, we know better!
*/
const wawActions: Record<CustomFunctions, (...args: any[]) => void> = {
screenshot: async (params: PageScreenshotOptions) => {
const screenshotBuffer = await page.screenshot({
...params, path: undefined,
});
await this.options.binaryCallback(screenshotBuffer, 'image/png');
},
enqueueLinks: async (selector: string) => {
const links: string[] = await page.locator(selector)
.evaluateAll(
// @ts-ignore
(elements) => elements.map((a) => a.href).filter((x) => x),
);
const context = page.context();
for (const link of links) {
// eslint-disable-next-line
this.concurrency.addJob(async () => {
try {
const newPage = await context.newPage();
2024-10-24 18:53:43 +05:30
await newPage.setViewportSize({ width: 900, height: 400 });
await newPage.goto(link);
await newPage.waitForLoadState('networkidle');
await this.runLoop(newPage, this.initializedWorkflow!);
} catch (e) {
// `runLoop` uses soft mode, so it recovers from it's own exceptions
// but newPage(), goto() and waitForLoadState() don't (and will kill
// the interpreter by throwing).
this.log(<Error>e, Level.ERROR);
}
});
}
await page.close();
},
scrape: async (selector?: string) => {
await this.ensureScriptsLoaded(page);
2024-08-03 02:20:59 +05:30
const scrapeResults: Record<string, string>[] = await page.evaluate((s) => window.scrape(s ?? null), selector);
await this.options.serializableCallback(scrapeResults);
2024-08-03 02:20:59 +05:30
},
2024-08-04 03:50:03 +05:30
scrapeSchema: async (schema: Record<string, { selector: string; tag: string, attribute: string; }>) => {
2024-08-03 02:20:59 +05:30
await this.ensureScriptsLoaded(page);
2024-08-12 06:35:00 +05:30
const scrapeResult = await page.evaluate((schemaObj) => window.scrapeSchema(schemaObj), schema);
await this.options.serializableCallback(scrapeResult);
2024-08-03 02:20:59 +05:30
},
2024-08-16 18:21:34 +05:30
scrapeList: async (config: { listSelector: string, fields: any, limit?: number, pagination: any }) => {
2024-08-06 23:20:10 +05:30
await this.ensureScriptsLoaded(page);
if (!config.pagination) {
const scrapeResults: Record<string, any>[] = await page.evaluate((cfg) => window.scrapeList(cfg), config);
await this.options.serializableCallback(scrapeResults);
} else {
const scrapeResults: Record<string, any>[] = await this.handlePagination(page, config);
await this.options.serializableCallback(scrapeResults);
}
2024-08-06 23:20:10 +05:30
},
2024-08-12 06:28:08 +05:30
scrapeListAuto: async (config: { listSelector: string }) => {
await this.ensureScriptsLoaded(page);
2024-08-12 06:35:00 +05:30
2024-08-12 06:28:08 +05:30
const scrapeResults: { selector: string, innerText: string }[] = await page.evaluate((listSelector) => {
return window.scrapeListAuto(listSelector);
}, config.listSelector);
2024-08-12 06:35:00 +05:30
2024-08-12 06:28:08 +05:30
await this.options.serializableCallback(scrapeResults);
},
scroll: async (pages?: number) => {
await page.evaluate(async (pagesInternal) => {
for (let i = 1; i <= (pagesInternal ?? 1); i += 1) {
// @ts-ignore
window.scrollTo(0, window.scrollY + window.innerHeight);
}
}, pages ?? 1);
},
2024-08-12 06:35:00 +05:30
script: async (code: string) => {
const AsyncFunction: FunctionConstructor = Object.getPrototypeOf(
async () => { },
).constructor;
const x = new AsyncFunction('page', 'log', code);
await x(page, this.log);
},
2024-08-18 22:19:24 +05:30
flag: async () => new Promise((res) => {
this.emit('flag', page, res);
}),
};
for (const step of steps) {
this.log(`Launching ${step.action}`, Level.LOG);
if (step.action in wawActions) {
// "Arrayifying" here should not be needed (TS + syntax checker - only arrays; but why not)
const params = !step.args || Array.isArray(step.args) ? step.args : [step.args];
await wawActions[step.action as CustomFunctions](...(params ?? []));
} else {
// Implements the dot notation for the "method name" in the workflow
const levels = step.action.split('.');
const methodName = levels[levels.length - 1];
let invokee: any = page;
for (const level of levels.splice(0, levels.length - 1)) {
invokee = invokee[level];
}
if (!step.args || Array.isArray(step.args)) {
await (<any>invokee[methodName])(...(step.args ?? []));
} else {
await (<any>invokee[methodName])(step.args);
}
}
await new Promise((res) => { setTimeout(res, 500); });
}
}
2024-08-17 23:54:00 +05:30
private async handlePagination(page: Page, config: { listSelector: string, fields: any, limit?: number, pagination: any }) {
let allResults: Record<string, any>[] = [];
let previousHeight = 0;
// track unique items per page to avoid re-scraping
2024-08-20 23:15:33 +05:30
let scrapedItems: Set<string> = new Set<string>();
2024-08-18 22:19:24 +05:30
2024-08-17 23:54:00 +05:30
while (true) {
switch (config.pagination.type) {
case 'scrollDown':
2024-08-18 22:18:03 +05:30
await page.evaluate(() => window.scrollTo(0, document.body.scrollHeight));
2024-08-18 22:19:24 +05:30
await page.waitForTimeout(2000);
const currentHeight = await page.evaluate(() => document.body.scrollHeight);
if (currentHeight === previousHeight) {
2024-08-18 22:32:39 +05:30
const finalResults = await page.evaluate((cfg) => window.scrapeList(cfg), config);
allResults = allResults.concat(finalResults);
return allResults;
2024-08-18 22:19:24 +05:30
}
2024-08-18 22:19:24 +05:30
previousHeight = currentHeight;
2024-08-17 23:54:00 +05:30
break;
case 'scrollUp':
2024-09-20 20:26:51 +05:30
await page.evaluate(() => window.scrollTo(0, 0));
await page.waitForTimeout(2000);
2024-09-20 23:37:00 +05:30
2024-09-20 20:26:51 +05:30
const currentTopHeight = await page.evaluate(() => document.documentElement.scrollTop);
if (currentTopHeight === 0) {
2024-09-20 23:37:00 +05:30
const finalResults = await page.evaluate((cfg) => window.scrapeList(cfg), config);
allResults = allResults.concat(finalResults);
return allResults;
2024-09-20 20:26:51 +05:30
}
2024-09-20 23:37:00 +05:30
2024-09-20 20:26:51 +05:30
previousHeight = currentTopHeight;
2024-08-17 23:54:00 +05:30
break;
2024-08-20 23:15:33 +05:30
case 'clickNext':
const pageResults = await page.evaluate((cfg) => window.scrapeList(cfg), config);
// Filter out already scraped items
const newResults = pageResults.filter(item => {
const uniqueKey = JSON.stringify(item);
if (scrapedItems.has(uniqueKey)) return false; // Ignore if already scraped
scrapedItems.add(uniqueKey); // Mark as scraped
return true;
});
allResults = allResults.concat(newResults);
if (config.limit && allResults.length >= config.limit) {
return allResults.slice(0, config.limit);
}
const nextButton = await page.$(config.pagination.selector);
if (!nextButton) {
return allResults; // No more pages to scrape
}
await Promise.all([
nextButton.click(),
page.waitForNavigation({ waitUntil: 'networkidle' })
]);
await page.waitForTimeout(1000);
break;
2024-08-17 23:54:00 +05:30
case 'clickLoadMore':
2024-09-20 23:35:49 +05:30
while (true) {
2024-09-20 23:37:00 +05:30
const loadMoreButton = await page.$(config.pagination.selector);
if (!loadMoreButton) {
// No more "Load More" button, so scrape the remaining items
const finalResults = await page.evaluate((cfg) => window.scrapeList(cfg), config);
allResults = allResults.concat(finalResults);
return allResults;
}
// Click the 'Load More' button to load additional items
await loadMoreButton.click();
await page.waitForTimeout(2000); // Wait for new items to load
// After clicking 'Load More', scroll down to load more items
await page.evaluate(() => window.scrollTo(0, document.body.scrollHeight));
await page.waitForTimeout(2000);
// Check if more items are available
const currentHeight = await page.evaluate(() => document.body.scrollHeight);
if (currentHeight === previousHeight) {
// No more items loaded, return the scraped results
const finalResults = await page.evaluate((cfg) => window.scrapeList(cfg), config);
allResults = allResults.concat(finalResults);
return allResults;
}
previousHeight = currentHeight;
if (config.limit && allResults.length >= config.limit) {
// If limit is set and reached, return the limited results
allResults = allResults.slice(0, config.limit);
break;
}
2024-08-17 23:54:00 +05:30
}
break;
default:
2024-08-18 22:34:46 +05:30
const results = await page.evaluate((cfg) => window.scrapeList(cfg), config);
allResults = allResults.concat(results);
return allResults;
2024-08-17 23:54:00 +05:30
}
2024-08-18 22:19:24 +05:30
2024-08-18 22:46:10 +05:30
if (config.limit && allResults.length >= config.limit) {
allResults = allResults.slice(0, config.limit);
break;
2024-08-17 23:54:00 +05:30
}
}
2024-08-18 22:19:24 +05:30
2024-08-17 23:54:00 +05:30
return allResults;
}
private async runLoop(p: Page, workflow: Workflow) {
// apply ad-blocker to the current page
await this.applyAdBlocker(p);
const usedActions: string[] = [];
let lastAction = null;
let repeatCount = 0;
/**
* Enables the interpreter functionality for popup windows.
* User-requested concurrency should be entirely managed by the concurrency manager,
* e.g. via `enqueueLinks`.
*/
p.on('popup', (popup) => {
this.concurrency.addJob(() => this.runLoop(popup, workflow));
});
/* eslint no-constant-condition: ["warn", { "checkLoops": false }] */
while (true) {
// Checks whether the page was closed from outside,
// or the workflow execution has been stopped via `interpreter.stop()`
if (p.isClosed() || !this.stopper) {
return;
}
try {
await p.waitForLoadState();
} catch (e) {
await p.close();
return;
}
let pageState = {};
try {
pageState = await this.getState(p, workflow);
} catch (e: any) {
this.log('The browser has been closed.');
return;
}
if (this.options.debug) {
this.log(`Current state is: \n${JSON.stringify(pageState, null, 2)}`, Level.WARN);
}
const actionId = workflow.findIndex(
(step) => this.applicable(step.where, pageState, usedActions),
);
const action = workflow[actionId];
this.log(`Matched ${JSON.stringify(action?.where)}`, Level.LOG);
if (action) { // action is matched
if (this.options.debugChannel?.activeId) {
this.options.debugChannel.activeId(actionId);
}
repeatCount = action === lastAction ? repeatCount + 1 : 0;
if (this.options.maxRepeats && repeatCount >= this.options.maxRepeats) {
return;
}
lastAction = action;
try {
await this.carryOutSteps(p, action.what);
usedActions.push(action.id ?? 'undefined');
} catch (e) {
this.log(<Error>e, Level.ERROR);
}
} else {
//await this.disableAdBlocker(p);
return;
}
}
}
private async ensureScriptsLoaded(page: Page) {
2024-08-17 23:54:00 +05:30
const isScriptLoaded = await page.evaluate(() => typeof window.scrape === 'function' && typeof window.scrapeSchema === 'function' && typeof window.scrapeList === 'function' && typeof window.scrapeListAuto === 'function' && typeof window.scrollDown === 'function' && typeof window.scrollUp === 'function');
if (!isScriptLoaded) {
await page.addInitScript({ path: path.join(__dirname, 'browserSide', 'scraper.js') });
}
}
/**
* Spawns a browser context and runs given workflow.
* \
* Resolves after the playback is finished.
* @param {Page} [page] Page to run the workflow on.
* @param {ParamType} params Workflow specific, set of parameters
* for the `{$param: nameofparam}` fields.
*/
public async run(page: Page, params?: ParamType): Promise<void> {
if (this.stopper) {
throw new Error('This Interpreter is already running a workflow. To run another workflow, please, spawn another Interpreter.');
}
/**
* `this.workflow` with the parameters initialized.
*/
this.initializedWorkflow = Preprocessor.initWorkflow(this.workflow, params);
await this.ensureScriptsLoaded(page);
this.stopper = () => {
this.stopper = null;
};
this.concurrency.addJob(() => this.runLoop(page, this.initializedWorkflow!));
await this.concurrency.waitForCompletion();
this.stopper = null;
}
public async stop(): Promise<void> {
if (this.stopper) {
await this.stopper();
this.stopper = null;
} else {
throw new Error('Cannot stop, there is no running workflow!');
}
}
}