Skip to content

a0viedo/webcodex

Repository files navigation

webcodex

A set of functions to collect data and metrics from websites.

Architecture

The idea behind all functions are to act as thin layers to accommodate some input parameters, do some action and transform the result to satify some output (or to have a specific side-effect, like daily functions). Most of the functions are designed to focus on a single metric but there are few expections (like the color detection).

HTTP functions

Daily functions

These functions are persisting the metrics into DynamoDB and are triggered by an SNS message originated externally.

Project structure

There are two important types of functions included in this project: functions that run on a daily basis and functions that are triggered by HTTP requests. The goal is to be able to re-use as much shared logic as possible between these two different types. The utils module is where the shared logic between these two live. In order to bundle just the dependencies needed for a given function the project uses Webpack and defines the necessary imports on a per-function basis, passing these dependencies back to the utils module.

Metrics

Colors defined in the CSS

To get the data for this metric a request is made to the website and all the stylesheets are aggregated. After that, the following regular expression is used to match hexadecimal (e.g. both #AAA and `#AAAAAA), rgb and hsla definitions:

/(#([\da-f]{3}){1,2}|(rgb|hsl)a\((\d{1,3}%?,\s?){3}(1|0?\.\d+)\)|(rgb|hsl)\(\d{1,3}%?(,\s?\d{1,3}%?){2}\))/ig

Finally after the list of colors is obtained the list is sorted using color-sorter.

Color detection

This function uses a headless browser to capture a screenshot and utilizes imagemagick to extract predominant colors from a 50x50 matrix.

HTML

This function traverses the DOM tree and aggregates into DOM elements frequency and attributes frequency.

CSS

The function first tries to retrieve the CSS content of the website using axios and if that fails then retries using cURL with an specific set of headers (due to browser fringerprinting). After getting the whole CSS content of the website the function uses Project Wallace's css-analyzer to get the CSS metrics.

HTTP Headers

The purpose of this function is to collect the HTTP headers using cURL.

Lighthouse

The function spins a headless browser to then connect Lighthouse to it and run basic audits. After getting the results back, it uploads the results to Github's gists and returns the

Technologies detection

The function focuses on identifyng libraries or frameworks that are loaded in a website. To achieve this, the function runs a headless browser and collects all the <script> links after the website is fully loaded. Once the complete list of JS resources is obtained the function downloads every resource, concatenates them and finds matches with regular expressions.

Why not use cheerio for that? Some popular websites started modifying the <head> element dynamically adding JavaScript bundles on-the-fly, making the task of grouping all the JS resources while not "running" any JavaScript within cheerio very difficult.

Utility functions

Convert-images

The function is built to convert screenshot images from PNG format to webp (and crops the image too) so that an up-to-date screenshot for every designated website is easily accessed.

Optimize image

The function can receive an image url and will return the same image converted to webp format in a base64 encoding format.

Data store

The metrics collected daily are being stored in DynamoDB. Every update to DynamoDB is designed to replace data at the property level, in a way similar to what an update with :set in MongoDB would do.

Next steps

There is a low percentage of errors on current executions and although the project does not aim to be able to target 100% of the websites out there, it would be good to know why those failures happen and where is the limitation.

About

The set of serverless functions that powers webcodex.dev

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published