vodorod/parcer
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge pull request #266 from getmaxun/sel-fix

Browse Source 
feat: selection revamp
This commit is contained in:
Karishma Shukla
2024-12-17 21:54:15 +05:30
committed by GitHub
parent fa2d6099ce a9dc4c8f4c
commit c25975bf98
3 changed files with 141 additions and 96 deletions
Download Patch File Download Diff File Expand all files Collapse all files

93
maxun-core/src/browserSide/scraper.js
View File

@@ -265,41 +265,72 @@ function scrapableHeuristics(maxCountPerPage = 50, minArea = 20000, scrolls = 3,
const scrapedData = [];
while (scrapedData.length < limit) {
// Get all parent elements matching the listSelector
const parentElements = Array.from(document.querySelectorAll(listSelector));
let parentElements = Array.from(document.querySelectorAll(listSelector));
// If we only got one element or none, try a more generic approach
if (limit > 1 && parentElements.length <= 1) {
const [containerSelector, _] = listSelector.split('>').map(s => s.trim());
const container = document.querySelector(containerSelector);
if (container) {
const allChildren = Array.from(container.children);
const firstMatch = document.querySelector(listSelector);
if (firstMatch) {
// Get classes from the first matching element
const firstMatchClasses = Array.from(firstMatch.classList);
// Find similar elements by matching most of their classes
parentElements = allChildren.filter(element => {
const elementClasses = Array.from(element.classList);
// Iterate through each parent element
for (const parent of parentElements) {
if (scrapedData.length >= limit) break;
const record = {};
// For each field, select the corresponding element within the parent
for (const [label, { selector, attribute }] of Object.entries(fields)) {
const fieldElement = parent.querySelector(selector);
if (fieldElement) {
if (attribute === 'innerText') {
record[label] = fieldElement.innerText.trim();
} else if (attribute === 'innerHTML') {
record[label] = fieldElement.innerHTML.trim();
} else if (attribute === 'src') {
// Handle relative 'src' URLs
const src = fieldElement.getAttribute('src');
record[label] = src ? new URL(src, window.location.origin).href : null;
} else if (attribute === 'href') {
// Handle relative 'href' URLs
const href = fieldElement.getAttribute('href');
record[label] = href ? new URL(href, window.location.origin).href : null;
} else {
record[label] = fieldElement.getAttribute(attribute);
// Element should share at least 70% of classes with the first match
const commonClasses = firstMatchClasses.filter(cls =>
elementClasses.includes(cls));
return commonClasses.length >= Math.floor(firstMatchClasses.length * 0.7);
});
}
}
}
}
scrapedData.push(record);
}
// Iterate through each parent element
for (const parent of parentElements) {
if (scrapedData.length >= limit) break;
const record = {};
// For each field, select the corresponding element within the parent
for (const [label, { selector, attribute }] of Object.entries(fields)) {
const fieldElement = parent.querySelector(selector);
if (fieldElement) {
if (attribute === 'innerText') {
record[label] = fieldElement.innerText.trim();
} else if (attribute === 'innerHTML') {
record[label] = fieldElement.innerHTML.trim();
} else if (attribute === 'src') {
// Handle relative 'src' URLs
const src = fieldElement.getAttribute('src');
record[label] = src ? new URL(src, window.location.origin).href : null;
} else if (attribute === 'href') {
// Handle relative 'href' URLs
const href = fieldElement.getAttribute('href');
record[label] = href ? new URL(href, window.location.origin).href : null;
} else {
record[label] = fieldElement.getAttribute(attribute);
}
}
}
scrapedData.push(record);
}
// If we've processed all available elements and still haven't reached the limit,
// break to avoid infinite loop
if (parentElements.length === 0 || scrapedData.length >= parentElements.length) {
break;
}
}
return scrapedData
};
return scrapedData;
};
/**

8
server/src/workflow-management/classes/Generator.ts
View File

@@ -544,9 +544,9 @@ export class WorkflowGenerator {
* @returns {Promise<string|null>}
*/
private generateSelector = async (page: Page, coordinates: Coordinates, action: ActionType) => {
const elementInfo = await getElementInformation(page, coordinates, this.listSelector);
const elementInfo = await getElementInformation(page, coordinates, this.listSelector, this.getList);
const selectorBasedOnCustomAction = (this.getList === true)
? await getNonUniqueSelectors(page, coordinates)
? await getNonUniqueSelectors(page, coordinates, this.listSelector)
: await getSelectors(page, coordinates);
const bestSelector = getBestSelectorForAction(
@@ -572,9 +572,9 @@ export class WorkflowGenerator {
* @returns {Promise<void>}
*/
public generateDataForHighlighter = async (page: Page, coordinates: Coordinates) => {
const rect = await getRect(page, coordinates, this.listSelector);
const rect = await getRect(page, coordinates, this.listSelector, this.getList);
const displaySelector = await this.generateSelector(page, coordinates, ActionType.Click);
const elementInfo = await getElementInformation(page, coordinates, this.listSelector);
const elementInfo = await getElementInformation(page, coordinates, this.listSelector, this.getList);
if (rect) {
if (this.getList === true) {
if (this.listSelector !== '') {

136
server/src/workflow-management/selector.ts
View File

@@ -17,9 +17,10 @@ export const getElementInformation = async (
page: Page,
coordinates: Coordinates,
listSelector: string,
getList: boolean
) => {
try {
if (listSelector !== '') {
if (!getList || listSelector !== '') {
const elementInfo = await page.evaluate(
async ({ x, y }) => {
const el = document.elementFromPoint(x, y) as HTMLElement;
@@ -74,22 +75,10 @@ export const getElementInformation = async (
if (originalEl) {
let element = originalEl;
const containerTags = ['DIV', 'SECTION', 'ARTICLE', 'MAIN', 'HEADER', 'FOOTER', 'NAV', 'ASIDE',
'ADDRESS', 'BLOCKQUOTE', 'DETAILS', 'DIALOG', 'FIGURE', 'FIGCAPTION', 'MAIN', 'MARK', 'SUMMARY', 'TIME',
'TABLE', 'THEAD', 'TBODY', 'TFOOT', 'TR', 'TH', 'TD', 'CAPTION', 'COLGROUP', 'COL', 'FORM', 'FIELDSET',
'LEGEND', 'LABEL', 'INPUT', 'BUTTON', 'SELECT', 'DATALIST', 'OPTGROUP', 'OPTION', 'TEXTAREA', 'OUTPUT',
'PROGRESS', 'METER', 'DETAILS', 'SUMMARY', 'MENU', 'MENUITEM', 'MENUITEM', 'APPLET', 'EMBED', 'OBJECT',
'PARAM', 'VIDEO', 'AUDIO', 'SOURCE', 'TRACK', 'CANVAS', 'MAP', 'AREA', 'SVG', 'IFRAME', 'FRAME', 'FRAMESET',
'LI', 'UL', 'OL', 'DL', 'DT', 'DD', 'HR', 'P', 'PRE', 'LISTING', 'PLAINTEXT', 'A'
];
while (element.parentElement) {
const parentRect = element.parentElement.getBoundingClientRect();
const childRect = element.getBoundingClientRect();
if (!containerTags.includes(element.parentElement.tagName)) {
break;
}
const fullyContained =
parentRect.left <= childRect.left &&
parentRect.right >= childRect.right &&
@@ -167,9 +156,9 @@ export const getElementInformation = async (
* @category WorkflowManagement-Selectors
* @returns {Promise<Rectangle|undefined|null>}
*/
export const getRect = async (page: Page, coordinates: Coordinates, listSelector: string) => {
export const getRect = async (page: Page, coordinates: Coordinates, listSelector: string, getList: boolean) => {
try {
if (listSelector !== '') {
if (!getList || listSelector !== '') {
const rect = await page.evaluate(
async ({ x, y }) => {
const el = document.elementFromPoint(x, y) as HTMLElement;
@@ -202,22 +191,10 @@ export const getRect = async (page: Page, coordinates: Coordinates, listSelector
if (originalEl) {
let element = originalEl;
const containerTags = ['DIV', 'SECTION', 'ARTICLE', 'MAIN', 'HEADER', 'FOOTER', 'NAV', 'ASIDE',
'ADDRESS', 'BLOCKQUOTE', 'DETAILS', 'DIALOG', 'FIGURE', 'FIGCAPTION', 'MAIN', 'MARK', 'SUMMARY', 'TIME',
'TABLE', 'THEAD', 'TBODY', 'TFOOT', 'TR', 'TH', 'TD', 'CAPTION', 'COLGROUP', 'COL', 'FORM', 'FIELDSET',
'LEGEND', 'LABEL', 'INPUT', 'BUTTON', 'SELECT', 'DATALIST', 'OPTGROUP', 'OPTION', 'TEXTAREA', 'OUTPUT',
'PROGRESS', 'METER', 'DETAILS', 'SUMMARY', 'MENU', 'MENUITEM', 'MENUITEM', 'APPLET', 'EMBED', 'OBJECT',
'PARAM', 'VIDEO', 'AUDIO', 'SOURCE', 'TRACK', 'CANVAS', 'MAP', 'AREA', 'SVG', 'IFRAME', 'FRAME', 'FRAMESET',
'LI', 'UL', 'OL', 'DL', 'DT', 'DD', 'HR', 'P', 'PRE', 'LISTING', 'PLAINTEXT', 'A'
];
while (element.parentElement) {
const parentRect = element.parentElement.getBoundingClientRect();
const childRect = element.getBoundingClientRect();
if (!containerTags.includes(element.parentElement.tagName)) {
break;
}
const fullyContained =
parentRect.left <= childRect.left &&
parentRect.right >= childRect.right &&
@@ -875,8 +852,10 @@ interface SelectorResult {
* @returns {Promise<Selectors|null|undefined>}
*/
export const getNonUniqueSelectors = async (page: Page, coordinates: Coordinates): Promise<SelectorResult> => {
export const getNonUniqueSelectors = async (page: Page, coordinates: Coordinates, listSelector: string): Promise<SelectorResult> => {
try {
if (!listSelector) {
console.log(`NON UNIQUE: MODE 1`)
const selectors = await page.evaluate(({ x, y }: { x: number, y: number }) => {
function getNonUniqueSelector(element: HTMLElement): string {
let selector = element.tagName.toLowerCase();
@@ -914,47 +893,82 @@ export const getNonUniqueSelectors = async (page: Page, coordinates: Coordinates
let element = originalEl;
const containerTags = ['DIV', 'SECTION', 'ARTICLE', 'MAIN', 'HEADER', 'FOOTER', 'NAV', 'ASIDE',
'ADDRESS', 'BLOCKQUOTE', 'DETAILS', 'DIALOG', 'FIGURE', 'FIGCAPTION', 'MAIN', 'MARK', 'SUMMARY', 'TIME',
'TABLE', 'THEAD', 'TBODY', 'TFOOT', 'TR', 'TH', 'TD', 'CAPTION', 'COLGROUP', 'COL', 'FORM', 'FIELDSET',
'LEGEND', 'LABEL', 'INPUT', 'BUTTON', 'SELECT', 'DATALIST', 'OPTGROUP', 'OPTION', 'TEXTAREA', 'OUTPUT',
'PROGRESS', 'METER', 'DETAILS', 'SUMMARY', 'MENU', 'MENUITEM', 'MENUITEM', 'APPLET', 'EMBED', 'OBJECT',
'PARAM', 'VIDEO', 'AUDIO', 'SOURCE', 'TRACK', 'CANVAS', 'MAP', 'AREA', 'SVG', 'IFRAME', 'FRAME', 'FRAMESET',
'LI', 'UL', 'OL', 'DL', 'DT', 'DD', 'HR', 'P', 'PRE', 'LISTING', 'PLAINTEXT', 'A'
];
while (element.parentElement) {
const parentRect = element.parentElement.getBoundingClientRect();
const childRect = element.getBoundingClientRect();
if (!containerTags.includes(element.parentElement.tagName)) {
break;
// if (listSelector === '') {
while (element.parentElement) {
const parentRect = element.parentElement.getBoundingClientRect();
const childRect = element.getBoundingClientRect();
const fullyContained =
parentRect.left <= childRect.left &&
parentRect.right >= childRect.right &&
parentRect.top <= childRect.top &&
parentRect.bottom >= childRect.bottom;
const significantOverlap =
(childRect.width * childRect.height) /
(parentRect.width * parentRect.height) > 0.5;
if (fullyContained && significantOverlap) {
element = element.parentElement;
} else {
break;
}
}
const fullyContained =
parentRect.left <= childRect.left &&
parentRect.right >= childRect.right &&
parentRect.top <= childRect.top &&
parentRect.bottom >= childRect.bottom;
const significantOverlap =
(childRect.width * childRect.height) /
(parentRect.width * parentRect.height) > 0.5;
if (fullyContained && significantOverlap) {
element = element.parentElement;
} else {
break;
}
}
// }
const generalSelector = getSelectorPath(element);
return {
generalSelector,
};
}, coordinates);
return selectors || { generalSelector: '' };
} else {
console.log(`NON UNIQUE: MODE 2`)
const selectors = await page.evaluate(({ x, y }: { x: number, y: number }) => {
function getNonUniqueSelector(element: HTMLElement): string {
let selector = element.tagName.toLowerCase();
if (element.className) {
const classes = element.className.split(/\s+/).filter((cls: string) => Boolean(cls));
if (classes.length > 0) {
const validClasses = classes.filter((cls: string) => !cls.startsWith('!') && !cls.includes(':'));
if (validClasses.length > 0) {
selector += '.' + validClasses.map(cls => CSS.escape(cls)).join('.');
}
}
}
return selector;
}
function getSelectorPath(element: HTMLElement | null): string {
const path: string[] = [];
let depth = 0;
const maxDepth = 2;
while (element && element !== document.body && depth < maxDepth) {
const selector = getNonUniqueSelector(element);
path.unshift(selector);
element = element.parentElement;
depth++;
}
return path.join(' > ');
}
const originalEl = document.elementFromPoint(x, y) as HTMLElement;
if (!originalEl) return null;
let element = originalEl;
const generalSelector = getSelectorPath(element);
return {
generalSelector,
};
}, coordinates);
return selectors || { generalSelector: '' };
}
} catch (error) {
console.error('Error in getNonUniqueSelectors:', error);
return { generalSelector: '' };