Stack Scraper is a system for efficiently scraping information from complex web sites in a repeatable way, exporting directly to a data store.
Stack Scraper is good at collecting lots of semi-structured data from complicated, or even poorly-written, web sites in a repeatable manner.
- Scraping operations can be paused and resumed at a later time.
- Fault tolerant.
- No one-to-one relationship between URLs and data collected. Multiple sources of data can be collected from a single page and the ID of the data can be handled arbitrarily (for example, the ID for a page could actually be the name of an image on the page, or the MD5 of that image, or something else entirely).
- Data for a single record can be collected, and compiled, from multiple, consecutive web pages. For example, let's say some data is on a page and then more data is within a popup. Both of those pages can be scraped and be combined into a single record.
- The process for crawling, downloading, extracting data, and processing the data are all de-coupled. They can all be run back-to-back, or one-at-a-time, or even repeatedly.
example directory for a full sample scraper.
Stack Scraper provides the code to write a simple command-line application for downloading semi-structured data from complex web sites. However you'll need to take a number of things into consideration when you're building your
stack-scraper implementation, namely:
- Command-line Interface The implementation of the command-line utilty and where various utility files will be located.
- Scrapers An implementation of a basic scraper.
- File System Where downloaded files (html, images, etc.) will live.
- Datastore and Data Models Where extracted data and scrape logs will be stored, and how.
- Post-Processors If any post-processing on the extracted data will be completed and how to do it.
type: Type of scraper to load (e.g. 'images' or 'artists').
source: The name of the source to download (e.g. 'ndl' or '*').
--scrape: Scrape and process the results from the already-downloaded pages.
--process: Process the results from the already-downloaded pages.
--reset: Don't resume from where the last scrape left off.
--delete: Delete all the data associated with the particular source.
--debug: Output additional debugging information.
--test: Test scraping and extraction of a source.
--test-url: Test extraction against a specified URL.
rootDataDir(String): A full path to the root directory where downloaded data is stored. (See "File System" for more information.)
scrapersDir(String): A full path to the directory where scraper
.jsfiles are stored. (See "Scrapers" for more information.)
directories(Object, optional): A key-value set of names of directories paired with the relative path to the directory. These directories will be created inside the individual source directory inside the
rootDataDir. (See "File System" for more information.)
model(Function): A function representing the model in which extracted data will be stored. (See "Datastore and Data Models" for more information.)
logModel(Function): A function representing the log model for storing information about an in-progress site scrape. (See "Datastore and Data Models" for more information.)
postProcessors(Object, optional): An object whose keys are the names of model properties which should be processed and values are functions through which the data will be processed. (See "Post-Processors" for more information.)
Datastore and Data Models
MongoDB + Mongoose
dbFind(filter:Object, callback) dbFindById(id:String, callback) dbSave(data:Object, callback) dbUpdate(data:Object, newData:Object, callback) dbRemove(filter:Object, callback) dbLog(data:Object, callback) dbStreamLog(filter:Object) -> Stream dbRemoveLog(filter:Object, callback)