ssc-codegen 0.1.4

generate selector schemas classes from yaml config and DSL-lang script

Selector schema codegen

RUSSIAN ENGLISH

ssc_codegen - generator of parsers for various programming languages (for html priority) using yaml-DSL configurations with built-in declarative language.

Designed to port parsers to various programming languages

Minimum requirements for the target portable language:

• include regular expressions
• include css/xpath libs for more supports queries (nth-child, ...)
• basic methods for works with strings (format string, trim/left trim/right trim/split/replace)

Recommendations

• usage css selector: they can be guaranteed converted to xpath
• there is an xpath to css converter for simple queries without guarantees of functionality. For example, in css there is no analogue of contains from xpath, etc.

Schematic representation of generator operation

Syntax specs

Language features

• DSL (Domain-Specific Language), declarative (no assignment, arithmetic, priority operations)
• Minimalistic syntax for working with selectors, regular expressions and simple string operations
• All methods take one argument as input and it is always selector-like type
• 4 types
• Regular expression syntax is like in python. For maximum compatibility, use for example [0-9] instead of \d
• Empty lines and comments (//) are ignored by the parser.

Types description

There are 4 data types for this scripting language

Тип	Описание
SELECTOR	class instance (Document, Element) from which css/xpath selectors are called. Always the first argument
SELECTOR_ARRAY	representation of a list of nodes of all found elements from the SELECTOR instance
TEXT	string
ARRAY	array of strings

Description of directives

• statements are separated by line indentation \n
• All string arguments are specified with double " quotes.
• Space are ignored

Operator	Arguments	Description	Return type value	Example
default	""	Default value if an error occurred during parsing. Listed first	-	default "empty"
xpath	""	xpath selector, returns the first value found	SELECTOR	xpath "//title"
xpathAll	""	xpath selector, returns all values	SELECTOR	xpathAll "//div"
css	""	css selector, returns the first value found	SELECTOR	css "title"
cssAll	""	css selector, returns all values	SELECTOR	cssAll "div > a"
attr	""	get tag(s). Called after xpath/xpathAll/css/cssAll	TEXT/ARRAY	attr "href"
text		get the text inside the tag. Called after xpath/xpathAll/css/cssAll. Can be called first to completely convert a SELECTOR object to TEXT	TEXT/ARRAY	text
raw		get the raw tag as text. Called after xpath/xpathAll/css/cssAll	TEXT/ARRAY	raw
re	""	regular expression. Returns the first element found. Argument must be TEXT	TEXT	re "(\d+)"
reAll	""	regular expression. Returns all found elements. Argument must be TEXT	ARRAY	reAll "(\d+)"
reSub	"" ""	Replacement by regular expression. Argument must be TEXT	TEXT	reSub "(\d+)" "digit(lol)"
strip	""	Removes the given string LEFT and RIGHT. Argument must be TEXT	TEXT	strip "\n"
lstrip	""	Deletes the specified line from the LEFT. Argument must be TEXT	TEXT	lstrip " "
rstrip	""	Deletes the specified row on the RIGHT. Argument must be TEXT	TEXT	rstrip " "
format	""	Format string. Specify a substitution argument using the {{}} operator. Argument must be TEXT	TEXT	format "spam {{}} egg"
split	""	Splitting a line. If count = -1 or not transmitted, divide by the maximum available. Argument must be TEXT	ARRAY	split ", "
replace	"" ""	String replacement. If count = -1 or not passed, replace it with the maximum available one. Argument must be TEXT	ARRAY	split ", "
limit		Maximum number of elements	ARRAY	limit 50
index		Take element by index. Argument must be ARRAY	TEXT	index 1
first		index 1 alias	TEXT	first
last		index -1 alias	TEXT	last
join	""	Collects ARRAY into a string. Argument must be ARRAY	TEXT	join ", "
ret		Tell the translator to return a value. Automatically added if not specified in the script		ret
noRet	""	Tell the translator not to return anything. Added for document pre-validation		noRet
//	...	One line comment. Ignored by the final code generator		// this is comment line

Токены валидации

The following commands are needed to pre-validate the input document using assert and they do not change final and intermediate values.

In this DSL language there are no boolean, null types, so if the result is false it will throw an error like AssertionError.

The operators accept SELECTOR:

• assertCss
• assertXpath

All other operators accept TEXT:

Оператор	Описание	Пример
assertEqual	Full string comparison (==) (case sensitive)	assertEqual "lorem upsum dolor"
assertContains	Comparison by presence of part of a string in TEXT	assertContains "sum"
assertStarts	Comparison based on the presence of part of a string at the beginning of TEXT	assertStarts "lorem"
assertEnds	Comparison based on the presence of part of a string at the end of TEXT	assertEnds "dolor"
assertMatch	Compare TEXT by regular expression	assertMatch "lorem \w+ dolor"
assertCss	Checking the validity of the query in SELECTOR.	assertCss "head > title"
assertXpath	Checking the validity of the query in SELECTOR.	assertXpath "//head/title"

Example code generation

// set default value if parse process is failing
xpath "//title"
text
format "Cool title: {{}}"

generated python equivalent code:

from parsel import Selector


def dummy_parse(part: Selector):
    val_0 = part.xpath('//title')
    val_1 = val_0.xpath('/text()').get()
    val_2 = "Cool title: {}".format(val_1)
    return val_2

generated dart equivalent code:

import 'package:html/parser.dart' as html;

dummy_parse(part){
    var val_0 = part.querySelector('title');
    String val_1 = val_0?.text ?? "";
    var val_2 = "Cool title: $val_1";
    return val_2;
}

add default value:

// set default value if parse process is failing
default "spam egg"
xpath "//title"
text
format "Cool title: {{}}"

from parsel import Selector


def dummy_parse(part: Selector):
    try:  
      val_1 = part.xpath('//title')
      val_2 = val_1.xpath('/text()').get()
      val_3 = "Cool title: {}".format(val_2)
      return val_3
    except Exception:
        return "spam egg"

import 'package:html/parser.dart' as html;

dummy_parse(html.Document part){
  try{
    var val_0 = part.querySelector('title');
    String val_1 = val_0?.text ?? "";
    var val_2 = "Cool title: $val_1";
    return val_2;
  } catch (e){
    return "spam egg";
  }
    
}

add assert validator

// not null check operation
assertCss "head > title"
xpath "//title"
text
format "Cool title: {{}}"

from parsel import Selector


def dummy_parse(part: Selector):
    assert part.css("head > title")
    val_1 = part.xpath('//title')
    val_2 = val_1.xpath('/text()').get()
    val_3 = "Cool title: {}".format(val_2)
    return val_3

import 'package:html/parser.dart' as html;

dummy_parse(html.Document part){
    assert(part.querySelector('title') != null);
    var val_0 = part.querySelector('title');
    String val_1 = val_0?.text ?? "";
    var val_2 = "Cool title: $val_1";
    return val_2;
}

yaml config

An example of the structure of the generated parser class:

• selector - Selector/Document instance, initialized using document
• _aliases - remapping keys for the view() method
• _viewKeys - output keys for the view() method
• _cachedResult - cache of obtained values from the parse() method
• parse() - launching the parser
• view() - getting the received values
• _preValidate() - an optional method of preliminary validation of the input document according to the rules from the configuration. If the result is false/null, it throws AssertError
• _partDocument() - an optional method of dividing a document into parts using a given selector. Useful, for example, for obtaining elements of the same type (product cards, etc.)
• _parseA, _parseB, _parseC, ... - automatically generated parser methods for each key (A,B,C) according to the rules from the configuration

Usage pseudocode example:

document = ... // extracted html document
instance = Klass(document)
instance.parse()
print(instance.view())

Example configuration file, see in examples

dev

DEV

TODO

• generated schemas checksum
• filter operations (?)
• constants support
• more languages, libs support
• codegen optimizations (usage SELECTOR fluent interfaces, one-line code generation)
• css/xpath analyzer in pre-generate step
• css/xpath patches (for example, if css selectors in target language not support :nth-child operation?)
• translate regex expressions. Eg: \d to [0-9]
• string methods: title, upper, lower, capitalize or any useful