ssc_codegen

Selector Schema codegen

Introduction

ssc-gen - a python-based DSL to describe parsers for html documents, which is translated into a standalone parsing module

For a better experience using this library, you should know:

• HTML CSS selectors (CSS3 standard min), Xpath
• regular expressions (PCRE)

Project solving next problems:

• designed for SSR (server-side-render) html pages parsers, NOT FOR REST-API, GRAPHQL ENDPOINTS
• decrease boilerplate code
• generates independent modules from the project that can be reused.
• generates docstring documentation and the signature of the parser output.
• for a better IDE experience, generates a typedefs, type annotations (if the target programming language supports it).
• support annotation and parsing of JSON-like strings from a document
• AST API codegen for developing a converter for parsing

Support converters

Current support converters

Language	HTML parser lib + dependencies	XPath	CSS3	CSS4	Generated annotations, types, structs	formatter dependency
Python (3.8-3.13)	bs4, lxml ( typing_extensions if py < 3.10 )	N	Y	Y	TypedDict1, list, dict	ruff
...	parsel ( typing_extensions if py < 3.10 )	Y	Y	N	...	...
...	selectolax (lexbor) ( typing_extensions if py < 3.10 )	N	Y	N	...	...
...	lxml ( typing_extensions if py < 3.10 )	Y	Y	N	...	...
js (ES6)2	pure (firefox/chrome extension/nodejs)	Y	Y	Y	JSDoc	prettier
go (1.10+) (UNSTABLE)	goquery, gjson (4)	N	Y	N	struct(+json anchors), array, map	gofmt
lua (5.2+), luajit(2+) (UNSTABLE)5	lua-htmlparser, lrexlib(opt), dkjson	N	Y	N	EmmyLua	LuaFormatter

• CSS3 means support next selectors:

  • basic: (tag, .class, #id, tag1,tag2)
  • combined: (div p, ul > li, h2 +p, title ~head)
  • attribute: (a[href], input[type='text'], a[href*='...'], ...)
  • CSS3 pseudo classes: (:nth-child(n), :first-child, :last-child)

• CSS4 means support next selectors:

  • :nth-of-type(), :where(), :is(), :not() etc

• 1this annotation type was deliberately chosen as a compromise reasons: Python has many ways of serialization: namedtuple, dataclass, attrs, pydantic, msgspec, etc

  • TypedDict is like a build-in dict, but with IDE and linter hint support, and you can easily implement an adapter for the required structure.

• 2ES8 standart required if needed use PCRE re.S | re.DOTALL flag

• 3js exclude build-in serialization methods, used standard Array and Map types. Focus on the singanutur documentation!

• 4golang has not been tested much, there may be issues

• formatter dependency - optional dependency for prettify and fix codestyle

• 5lua

  • Experimental Research PoC, performance and stability are not guaranteed
  • Priority on generation to pure lua without C-libs dependencies. using mva/htmlparser and dhkolf/dkjson
  • Translates unsupported CSS3 selectors into the equivalent in the form of function calls:
    • for example, div +p is equivalent to CssExt.combine_plus(root:select("div"), "p")
  • Translates PCRE regex to string pattern matching (with restrictions) for more information in lua_re_compat.py

Limitations

For maximum portability of the configuration to the target language:

• If possible, use CSS selectors: they are guaranteed to be converted to XPATH
• Unlike javascript, most html parse libs implement CSS3 selectors standard. They may not fully implement the functionality! Check the html parser lib documentation aboud CSS selectors before implement code. Examples:
  • Several libs not support + operations (eg: selectolax(modest), dart.universal_html)
  • For research purpose, lua_htmlparser include converter for unsupported CSS3 query syntax

• HTML parser libs maybe not supports attribute selectors: *=, ~=, |=, ^=, $=
• Several libs not support pseudo classes (eg: standard dart.html lib miss this feature).

Getting started

ssc_gen required python 3.10 version or higher

Install

pip:

pip install ssc_codegen

uv:

uv pip install ssc_codegen

Example

Create a file schema.py with:

from ssc_codegen import ItemSchema, D

class HelloWorld(ItemSchema):
    title = D().css('title').text()
    a_hrefs = D().css_all('a').attr('href')

try it in cli

[!note] this tools developed for testing purposes, not for web-scraping tasks

eval from file

Download any html file and pass as argument:

ssc-gen parse-from-file index.html -t schema.py:HelloWorld

Short options descriptions:

• -t --target - config schema file and class from where to start the parser

send GET request to url and parse response

ssc-gen parse-from-url https://example.com -t schema.py:HelloWorld

send request via Chromium browser (CDP protocol)

ssc-gen parse-from-chrome https://example.com -t schema.py:HelloWorld

[!note] if script cannot found chrome executable - provide it manually:

ssc-gen parse-from-chrome https://example.com -t schema.py:HelloWorld -sc /usr/bin/chromium

Convert to code

Convert to code for use in projects:

[!note] for example, used js: it can be fast test in developer console

ssc-gen js schema.py -o .

Code output looks like this:

// autogenerated by ssc-gen DO NOT_EDIT
/***
 *
 * {
 *     "title": "String",
 *     "a_hrefs": "Array<String>"
 * }*/
class HelloWorld {
  constructor(doc) {
    if (typeof doc === "string") {
      this._doc = new DOMParser().parseFromString(doc, "text/html");
    } else if (doc instanceof Document || doc instanceof Element) {
      this._doc = doc;
    } else {
      throw new Error("Invalid input: Expected a Document, Element, or string");
    }
  }

  _parseTitle(v) {
    let v0 = v.querySelector("title");
    return typeof v0.textContent === "undefined"
      ? v0.documentElement.textContent
      : v0.textContent;
  }

  _parseAHrefs(v) {
    let v0 = Array.from(v.querySelectorAll("a"));
    return v0.map((e) => e.getAttribute("href"));
  }

  parse() {
    return {
      title: this._parseTitle(this._doc),
      a_hrefs: this._parseAHrefs(this._doc),
    };
  }
}

Copy code output and past to developer console:

Print output:

alert(JSON.stringify(new HelloWorld(document).parse()));

You can use any html source:

• parse from html files
• parse from http responses
• parse from browsers: playwright, selenium, chrome-cdp, etc.
• call curl in shell and parse STDIN
• use in STDIN pipelines with third-party tools like projectdiscovery/httpx

See also

• Brief about css selectors and regular expressions
• Explain short document on how to understand DSL syntax
• LLM experimental prompt for generate code
• Explain short note how to explain and read sscgen schema configs
• Quickstart about css selectors and regular expressions.
• Tutorial basic usage ssc-gen
• AST reference about generation code from AST