Documentation

Table catalog: extract product rows (Python)

Read a table-style catalog and produce familiar product fields. This example uses page element selectors rather than column numbers.

Choose Python, paste the complete source from this article and set the run parameters below. The source for this collection is ScrapingCourse. This walkthrough applies to the specified demonstration page.

Example settings

{
  "path": "/table-parsing",
  "mode": "http",
  "expected": 15
}

path selects the page, mode selects an algorithm branch, and expected sets the control count for this example. It is not a plan limit. When the source dataset changes, check the expected count against it.

How the algorithm works

Open /table-parsing with the http mode. A normal request is sufficient to obtain the original markup.
Check the response and pass its HTML to the shared collector.
Select .product-item, li.product elements and read the name, price, link and image within each. The table presentation does not change how fields are located.
Convert the price text to a number with SDK text tools. Use the product link as its identifier, falling back to its name when the link is absent.
Check for a next-page link with the same HTTP algorithm. If none is present, finish reading.
Compare the unique row count with expected and return records containing name, price and image.

Products are processed by collect(html). Read the name first, skip empty names, and use the card link as the key, falling back to the name when necessary. seen prevents the same key being emitted twice. This matters when a later HTML read includes products already collected.

Details that matter

This example parses a particular HTML structure. For another catalog, map its fields to the collector selectors first; the visible column order is not used.

Python details

The Python SDK is sequential here: no async or await is needed. Method names use snake_case: without_cache, wait_for, wait_for_response. The nested collect uses a shared key set and returns the current HTML card count. finally calls page.close().

Expected output

The configured demonstration dataset expects 15 unique records. The final check compares the actual unique key count with expected.

Field	Contents
`name`	Product name with leading and trailing whitespace removed.
`price`	Price converted from text to a number.
`image`	Image URL read from the card src attribute.

Complete scenario source

The source is shared by this collection. This example runs its http branch using the parameters above. Copy the complete block into the Python editor; no fixed sign-in credentials or service keys need to be added.

import re


def run(ctx):
    origin = "https://www.scrapingcourse.com"
    path = str(ctx.param("path", "/pagination"))
    mode = str(ctx.param("mode", "http"))
    auto = ctx.param("auto_captcha", True) is not False
    seen = set()

    def collect(html):
        dom = ctx.dom(html)
        for card in dom.qa(".product-item, li.product"):
            title = card.q(".product-name, .woocommerce-loop-product__title")
            name = title.text().strip() if title else ""
            link = card.q("a")
            key = (link.attr("href") if link else "") or name
            if not name or key in seen:
                continue
            price = card.q(".product-price, .price")
            image = card.q("img")
            seen.add(key)
            ctx.emit({"name": name, "price": ctx.text().money(price.text() if price else ""),
                      "image": image.attr("src") if image else ""}, key)
        return len(dom.qa(".product-item, li.product"))

    if mode == "http":
        url = origin + path
        visited = set()
        while url and len(visited) < 50:
            if url in visited:
                raise ValueError("Pagination returned an already processed page")
            visited.add(url)
            response = ctx.http().without_cache().get(url, auto)
            if not response.ok():
                raise ValueError("The source could not be opened: " + str(response.status))
            html = response.body
            collect(html)
            next_url = ctx.dom(html).attr('a[rel="next"], a.next.page-numbers', "href")
            url = ctx.url().absolutize(url, next_url) if next_url else None
        if url:
            raise ValueError("The pagination safety limit was reached")
    else:
        page = ctx.browser().open(origin + path, ["*/ajax/products*"], {}, auto)
        try:
            if mode == "login":
                dom = ctx.dom(page.content())
                hint = next((p.text() for p in dom.qa("p") if "Demo Credentials:" in p.text()), "")
                password = re.search(r"Password:\s*([^\s|]+)", hint, re.I)
                email_match = re.search(r"Email:\s*([^\s|]+@[^\s|]+)", hint, re.I)
                email = email_match.group(1) if email_match else ""
                encoded = dom.attr("[data-cfemail]", "data-cfemail")
                if not email and encoded and len(encoded) > 2:
                    key = int(encoded[:2], 16)
                    email = "".join(chr(int(encoded[i:i + 2], 16) ^ key) for i in range(2, len(encoded), 2))
                if not email or not password:
                    raise ValueError("The page does not contain demo credentials")
                page.type("#email", email)
                page.type("#password", password.group(1))
                page.click('button[type="submit"]')
            if mode == "challenge":
                text = re.sub(r"\s+", " ", ctx.dom(page.content()).text("body") or "")
                if not re.search(r"You (?:successfully )?bypassed|successfully bypassed|Congratulations", text, re.I):
                    raise ValueError("The expected page content was not received")
                ctx.emit({"name": path, "passed": True}, path)
            else:
                page.wait_for(".product-name:not(:empty)", 15000)
                if mode == "button":
                    complete = False
                    for step in range(30):
                        html = page.content()
                        before = ctx.dom(html)
                        collect(html)
                        button = before.q("#load-more-btn")
                        if not button or re.search(r"display:\s*none", button.attr("style") or ""):
                            complete = True
                            break
                        offset = button.attr("data-offset")
                        count = len(before.qa(".product-item"))
                        page.click("#load-more-btn")
                        response = page.wait_for_response("*offset=" + offset, 15000)
                        if not response or not 200 <= response["status"] < 300 or response["truncated"] or response["error"]:
                            raise ValueError("The next batch could not be loaded")
                        if not (response.get("body") or "").strip():
                            complete = True
                            break
                        page.wait_for(".product-item:nth-child(" + str(count + 1) + ")")
                    if not complete:
                        raise ValueError("The batch loading safety limit was reached")
                if mode == "scroll":
                    stable, previous = 0, 0
                    for step in range(40):
                        page.scroll(3, 400)
                        count = collect(page.content())
                        stable = stable + 1 if count == previous else 0
                        previous = count
                        if stable >= 3:
                            break
                    if stable < 3:
                        raise ValueError("The feed is still loading: increase the scroll limit")
                collect(page.content())
        finally:
            page.close()
    expected = int(ctx.param("expected", 1))
    if mode != "challenge" and len(seen) != expected:
        raise ValueError("Incomplete result: " + str(len(seen)) + " instead of " + str(expected))