The crawler was incorrectly using the request URL as the base URL when resolving relative links. This caused problems when encountering redirects.
For example if we fetch /log, redirecting to /log/ and find links to foo/, and bar/; these would resolve to /foo and /bar, and not /log/foo and /log/bar.
Discovered feed URLs were given a double slash after their domain name in the DB. This will go away in the URL normalizer, so the URLs are still viable, but the commit fixes the issue regardless.
Feed discover is improved with by probing a few likely endpoints when no feed link tag is provided. To store the feed URLs, a sqlite database is added to each crawlset that stores a simple summary of the crawl job, including any feed URLs that have been discovered.
Solves issue #135
Also alter CrawledDocument to not require String parsing of the underlying byte[] data. This should reduce the number of large memory allocations quite significantly, hopefully reducing the GC churn a bit.
Some refactoring is still needed, but an dummy actor is in place and a process that crawls URLs from the livecapture service's RSS endpoints; that makes it all the way to being indexable.
Also be a bit smarter about pre-allocating queues and sets based on depth rather than the number of provided URLs, which was always zero outside of tests.
The crawl spec abstraction was used to upload lists of domains into the system for future crawling. This was fairly clunky, and it was difficult to understand what was going to be crawled.
Since a while back, a new domains listing view has been added to the control view that allows direct access to the domains table. This is much preferred and means the operator can directly manage domains without specs.
This commit removes the crawl spec abstraction from the code, and changes the GUI to direct to the domains list instead.
* Restructure the code to make a bit more sense
* Store full headers in crawl data
* Fix bug in retry-after header that assumed the timeout was in milliseconds, and then clamped it to a lower bound of 500ms, meaning this was almost always handled wrong
Turns out throttling to only 1 lock per domain means the crawler chokes hard on large hosting websites such as wordpress. Giving these a slightly larger allowance.
Refactoring keyword extraction to extract spans information.
Modifying the intermediate storage of converted data to use the new slop library, which is allows for easier storage of ad-hoc binary data like spans and positions.
This is a bit of a katamari damacy commit that ended up dragging along a bunch of other fairly tangentially related changes that are hard to break out into separate commits after the fact. Will push as-is to get back to being able to do more isolated work.
To let up the pressure on domains with lot sof subdomains such as substack, medium, neocities, etc. a per-domain mutex is added that will limit crawling of these domains to one thread at a time.
The revisit logic wasn't sufficiently dampening the recrawl rate for websites that largely have not changed.
Modified it to be more reactive to the degree to which the content has changed, while applying upper and lower limits depending on the size of the crawl set.
Look, this will make the git history look funny, but trimming unnecessary depth from the source tree is a very necessary sanity-preserving measure when dealing with a super-modularized codebase like this one.
While it makes the project configuration a bit less conventional, it will save you several clicks every time you jump between modules. Which you'll do a lot, because it's *modul*ar. The src/main/java convention makes a lot of sense for a non-modular project though. This ain't that.
