Loading [MathJax]/extensions/Safe.js

https://www.cars-data.com/en/hatchback-cars.html

import requests
from bs4 import BeautifulSoup

def get_html(url):
    resp = requests.get(url)
    return BeautifulSoup(resp.content, 'html.parser')

home_html = get_html("https://www.cars-data.com/")
brand_as = home_html.select(".brands_plus_most .row .col-5 .row.center div.col-2 a")
brand_links = [a["href"] for a in brand_as]

len(brand_links)

89

from ipywidgets import FloatProgress
from IPython.display import display

%%time
bar = FloatProgress(min=0, max=len(brand_links))
display(bar)

seriess_links = []
for brand_link in brand_links:
    brand_html = get_html(brand_link)
    series_as = brand_html.select("section.models div.col-4 a")
    series_links = [a["href"] for a in series_as]
    seriess_links += series_links
    bar.value += 1

CPU times: user 4.95 s, sys: 258 ms, total: 5.21 s
Wall time: 1min 12s

len(seriess_links)

1670

seriess_links = set(seriess_links)
len(seriess_links)

1670

%%time
bar = FloatProgress(min=0, max=len(seriess_links))
display(bar)

models_links = []
for series_link in seriess_links:
    series_html = get_html(series_link)
    model_as = series_html.select("section.models div.col-4 a")
    model_links = [a["href"] for a in model_as]
    models_links += model_links
    bar.value += 1

len(models_links)

models_links = set(models_links)
len(models_links)

list(models_links)[:3]

%%time
bar = FloatProgress(min=0, max=len(models_links))
display(bar)

variants_links = []
for model_link in models_links:
    model_html = get_html(model_link)
    variant_as = model_html.select("section.types div.col-8 div.row div.col-6 h2 a")
    variant_links = [a["href"] for a in variant_as]
    variants_links += variant_links
    bar.value += 1

len(variants_links)

variants_links = list(set(variants_links))
len(variants_links)

variants_links.remove("")

variants_links[:3]

from itertools import zip_longest
def grouper(iterable, n, fillvalue=None):
    """Collect data into fixed-length chunks or blocks
    https://docs.python.org/3/library/itertools.html#itertools-recipes
    """
    # grouper('ABCDEFG', 3, 'x') --> ABC DEF Gxx"
    args = [iter(iterable)] * n
    return zip_longest(*args, fillvalue=fillvalue)

%%time
bar = FloatProgress(min=0, max=len(variants_links))
display(bar)

datas = []
for variant_link in variants_links:
    variant_html = get_html(variant_link)
    data = {"url": variant_link}
    breadcrumb = variant_html.select("#breadcrumb")
    breadcrumbs = [el.text.strip() for el in variant_html.select("#breadcrumb a")]
    _, data["make"], data["model"], data["series"], data["variant"] = breadcrumbs
    
    dts = variant_html.select("dt")
    dds = variant_html.select("dd")
    
    for dt, dd in zip(dts, dds):
        data[dt.text.strip(":")] = dd.text
    datas.append(data)
    bar.value += 1

len(datas)

import pandas as pd

%%time
df = pd.DataFrame(datas)

df.iloc[0]

df.info(memory_usage="deep")

%%time
df.to_csv("card_data_tabs.csv", sep="\t", encoding="utf-8")

list(df.columns)