- A modern spider design that uses AnyIO and asyncio, yet it's very similar to Scrapy spiders API because it's the easiest design for users, and to make it easier for new users.
- Spiders can have multiple sessions per crawl, and users decide which session to use with each request.
- A scheduler system that uses heapq logic.
- The user can set the number of concurrent requests for a spider globally or per domain.
- The user can set a download delay to control the speed of the spider more.
- There's a global function that can be overridden to handle errors for all requests. (Similar to errback in scrapy).
- There's a spider argument to set the allowed domains for the spider to stay in.
- Each spider has a very detailed crawl stats that can be accessed right away from the code after the crawl finishes. Same case with scraped items.
- The whole spider as written as any other script and you just run it. No command-line arguments, and no need to run it from the terminal through the library like other known alternatives.
- Each spider has its own logger that forces sessions to use it.
- Each spider has functions to override that run before start and after close.
- There's a spider argument to set the logging level and another one to make the spider write to a log file.
- This is only the start. A lot more features are coming in the way.
- DynamicFetcher became 20% faster
- StealthyFetcher became 99% faster
- Scrapling size decreased
- Code became ~400 lines shorter
- Most importantly, scrapling is more stable and reliable now.
- Less confusing for new users.
- More...
Made the code shorter by an additional ~200 lines and easier to maintain in return for making the arguments autocompletion bad for shells that don't check for dynamic type hints like IPython.
Made the code shorter by about ~500 lines and easier to maintain in return for making the arguments autocompletion bad for shells that don't check for dynamic type hints like IPython.
Karim shoair