DEV Community

Cover image for How to build a CSS pre-processor (like SASS) from scratch
Abdur-Rahmaan Janhangeer
Abdur-Rahmaan Janhangeer

Posted on • Edited on • Originally published at pythonkitchen.com

How to build a CSS pre-processor (like SASS) from scratch

#python #compiler #showdev #css

If you are in web development, maybe you've heard of Sass, Less, Pug, Stylus etc. All these are pre-processors. In this tutorial we're going to build nothing less than a functional css pre-processor from scratch with variables and functions. This type of new language is called source-to-source compiled. If you are thrilled, i hope not to disappoint you.

Charting The Routes

First let’s see what we’ll be doing. We’ll call the language DotDot with extention .dot

When designing a new language, it’s good to spend sometimes on design. If you have something that really flows, you can start coding.

Here is a DotDot snippet:

x = 1
y = 5
x-y-z = 6
col-z = berry

@f{
    border: 10px solid black;
    border-radius: 50%;
}

.add{
    color:white;
    background-color: rgb(3, 4, 5);
}

#ad{
    color: rgb(x, 4, 5);
}

.a div a{
    color: ..col-z;
    @f
}
Enter fullscreen mode Exit fullscreen mode

which compiles to this:

.add{
    color: white;
    background-color: rgb(3,4,5);
}

#ad{
    color: rgb(1,4,5);
}

.a div a{
    color: berry;
    border: 10px solid black;
    border-radius: 50%;
}
Enter fullscreen mode Exit fullscreen mode

Dissecting Design

Let’s see what features we included

Variables 

x = 1
y = 5
x-y-z = 6
col-z = berry
Enter fullscreen mode Exit fullscreen mode

We see that

  • we don’t need to specify a specifier like $ or var in Js.
  • we can add – in the variable name

We can also call variables in function values

rgb(x, 4, 5);
Enter fullscreen mode Exit fullscreen mode

And in attribute values

color: ..col-z;
Enter fullscreen mode Exit fullscreen mode

Where .. denotes a variable call in attribute directly.

Functions 

@f{
    border: 10px solid black;
    border-radius: 50%;
}
Enter fullscreen mode Exit fullscreen mode

Functions are denoted by an @ sign. These are called as in the following case where it expands into the properties it contains.

.a div a{
    color: ..col-z;
    @f
}
Enter fullscreen mode Exit fullscreen mode

That is enough complexity to deal with. Let’s start!

import copy # we have just one import
Enter fullscreen mode Exit fullscreen mode

and we put the source as a variable

source = '''

x = 1
y = 5
x-y-z = 6
col-z = berry

@f{
    border: 10px solid black;
    border-radius: 50%;
}

.add{
    color:white;
    background-color: rgb(3, 4, 5);
}

#ad{
    color: rgb(x, 4, 5);
}

.a div a{
    color: ..col-z;
    @f
}
'''
Enter fullscreen mode Exit fullscreen mode

Defining Constants

We’ve bundled our constants in a class.

class SYM:
    LEFT_BRACE = '{'
    RIGHT_BRACE = '}'
    LEFT_ROUND = '('
    RIGHT_ROUND = ')'
    NEW_LINE = '\n'
    TAB = '    '
    COLON = ':'
    SEMI_COLON = ';'
    SPACE = ' '
    COMMA = ','
    EQUAL = '='
    UNION = '-'
    DOT = '.'
    AT = '@'
Enter fullscreen mode Exit fullscreen mode

We then define our keywords

KEYWORDS = (SYM.LEFT_BRACE, SYM.RIGHT_BRACE, SYM.NEW_LINE, SYM.TAB, SYM.COLON, 
    SYM.SEMI_COLON, SYM.SPACE, SYM.RIGHT_ROUND, SYM.LEFT_ROUND, SYM.COMMA, SYM.EQUAL)
Enter fullscreen mode Exit fullscreen mode

Building the Lexer

The code for the lexer is from this lexer tutorial. Please go over it if you feel the need to. Here we converted the code into a class with a get_lexeme method. The method gives us a list. The dotdot snippet above gets converted to the list held in lexemes variable below.

Here is our lexer class:

class Lexer:
    def __init__(self, source: str, KEYWORDS: list):
        self.white_space = SYM.SPACE
        self.KEYWORDS = KEYWORDS
        self.lexeme = ''
        self.lexemes = []
        self.string = source

    def get_lexemes(self) -> list:
        for i, char in enumerate(self.string):
            if char != self.white_space and char != SYM.NEW_LINE:
                self.lexeme += char  # adding a char each time
            if (i+1 < len(self.string)):  # prevents error
                if self.string[i+1] == self.white_space or self.string[i+1] in KEYWORDS or self.lexeme in KEYWORDS or self.string[i+1] == SYM.NEW_LINE: # if next char == ' '
                    if self.lexeme != '':
                        self.lexemes.append(self.lexeme)
                        self.lexeme = ''
        return self.lexemes
Enter fullscreen mode Exit fullscreen mode

Then we declare our basic variables:

v = Lexer(source, KEYWORDS)
lexemes = v.get_lexemes()

lexemes.append('') # appending an unused last element to properly dump all values
Enter fullscreen mode Exit fullscreen mode

lexemes is now equal to

['x', '=', '1', 'y', '=', '5', 'x-y-z', '=', '6', 'col-z', 
'=', 'berry', '@f', '{', 'border', ':', '10px', 'solid', 
'black', ';', 'border-radius', ':', '50%', ';', '}', '.add', 
'{', 'color', ':', 'white', ';', 'background-color', ':', 
'rgb', '(', '3', ',', '4', ',', '5', ')', ';', '}', '#ad', 
'{', 'color', ':', 'rgb','(', 'x', ',', '4', ',', '5', ')', 
';', '}', '.a', 'div', 'a', '{', 'color', ':', '..col-z', 
';', '@f', '}', '']
Enter fullscreen mode Exit fullscreen mode

With such separated data, it’s much easier to proceed!

The Concept of Memory

Next we define a dictionary to hold all our variables.

memory = {}
Enter fullscreen mode Exit fullscreen mode

where

x = 1
Enter fullscreen mode Exit fullscreen mode

will become this later on:

{'x':'1'}
Enter fullscreen mode Exit fullscreen mode

to retrieve it we’ll just do

memory['x']
Enter fullscreen mode Exit fullscreen mode

The Concept of Tree

We’ll have a dictionary called tree

tree = {}
Enter fullscreen mode Exit fullscreen mode

which will hold the converted code as

{
    '@f': {
        'border': '10px solid black', 
        'border-radius': '50%'
    }, 
    '.add': {
        'color':'white', 
        'background-color': 'rgb ( 3 , 4 , 5 )'
    }, '#ad': {
        'color': 'rgb ( x ,4 , 5 )'
    }, 
    '.a div a': {
        'color': '..col-z', 
        '@f': ''
    }
}
Enter fullscreen mode Exit fullscreen mode

Our next step will exactly be this: converting the lexemes list into this dictionary

Generating The Tree

To keep track of where we are, we’ll have a series of variables taking True/False values (on/off)

Setting Up Holders 

id_string = ''
last_id_string = ''
last_attribute = ''

current_attribute = ''
current_value = ''

len_lexemes = len(lexemes)
Enter fullscreen mode Exit fullscreen mode

id string will hold like #add, .x a div etc

the last_ variables just hold variables that are not emptied upon leaving the sub section

Setting Up Flags

We’ll have 3 flags.

One when we are starting a block, which will become true when encountering a { and false when encountering a }

An attribute is color in color:black;

The attribute ongoing will become true when passing over { and ; and will become false when passing over :

value_ongoing will become true when going over : and false when going over ;

block_ongoing = False
attribute_ongoing = False
value_ongoing = False
Enter fullscreen mode Exit fullscreen mode

We’ll start looping over the list and implement what we described of the flags above

for i, lex in enumerate(lexemes):

    if i+1 < len_lexemes:
        next_lexeme = lexemes[i+1]
    prev_lexeme = lexemes[i-1]

    if lex == SYM.LEFT_BRACE:
        block_ongoing = True
        attribute_ongoing = True
        continue
    elif lex == SYM.RIGHT_BRACE:
        block_ongoing = False
        statement_ongoing = False
        attribute_ongoing = False
        value_ongoing = False
        continue
    if lex == SYM.COLON:
        value_ongoing = True
        attribute_ongoing = False
        continue
    elif lex == SYM.SEMI_COLON:
        value_ongoing = False
        statement_ongoing = False
        attribute_ongoing = True
        continue
Enter fullscreen mode Exit fullscreen mode

Continue is needed as once we have activated a flag, we have no work to do, we move on.

Dealing with variables

To assign variables, we just wait for the = sign then continue.

Then when going over a variable name or a value we just prevent the loop from continuing by using continue

    if lex == SYM.EQUAL:
        memory[prev_lexeme] = next_lexeme
        continue
    elif next_lexeme == SYM.EQUAL or prev_lexeme == SYM.EQUAL:
        continue
Enter fullscreen mode Exit fullscreen mode

Where The Tree Builds

Now we’ll make use of the flags

    if not block_ongoing:
        id_string += lex + ' '
    elif block_ongoing:
        if id_string:
            tree[id_string.strip()] = {}
            last_id_string = id_string.strip()
            id_string = ''
Enter fullscreen mode Exit fullscreen mode

Here we are dealing with the block id, example #add in #add {}. We did

tree[id_string.strip()] = {}
Enter fullscreen mode Exit fullscreen mode

here is an example tree at this point

{
'.add': {}
}
Enter fullscreen mode Exit fullscreen mode

Using the same principle, we’ll do so for the attribute

    if attribute_ongoing:
        current_attribute += lex
    elif not attribute_ongoing:
        if current_attribute:
            tree[last_id_string][current_attribute] = ''
            last_attribute = current_attribute
            current_attribute = ''
Enter fullscreen mode Exit fullscreen mode

Here is an example tree at this point

{
'.add': {
        'color':''
    }
}
Enter fullscreen mode Exit fullscreen mode

and value

    if value_ongoing:
        current_value += lex + ' '
    elif not value_ongoing:
        if current_value:
            tree[last_id_string][last_attribute] = current_value.strip()
            last_value = current_value.strip()
            current_value = ''
Enter fullscreen mode Exit fullscreen mode

Here is an example tree at this point

{
'.add': {
        'color':'white'
    }
}
Enter fullscreen mode Exit fullscreen mode

That snippet ends our tree-building block

Replacing Values

Now let us see how to replace variable names in functions like rgb() and rgba()

def parseFunc(f):
    v = f.split(SYM.LEFT_ROUND)
    name = v[0]
    vals = v[1][:-1].replace(SYM.SPACE, '')
    values = vals.split(SYM.COMMA)
    for i, v in enumerate(values):
        if not v.isnumeric():
            values[i] = memory[v]
    return '{}({})'.format(name.strip(), ','.join(values))
Enter fullscreen mode Exit fullscreen mode

Here we replace rgb(x, 4, 5) to rgb(1, 4, 5) by replacing the x by 1 in the memory dictionary. To build a CSS pre-processor from scratch, we have to take all cases, which we’ll do after.

Transforming The Tree Into It’s Final Form

Now we need to pass over the tree once again and expand functions, replace values, including the use of our defined function above.

Technically we can’t change a dictionary while iterating over it. We have to copy it to break references. We have to deepcopy it.

ntree = copy.deepcopy(tree)
Enter fullscreen mode Exit fullscreen mode

then we iterate over:

for block_name in ntree:
    properties = ntree[block_name]
    if block_name[0] == SYM.AT:
        continue
Enter fullscreen mode Exit fullscreen mode

Now, if we get the first symbol of the id as a @, we just skip it as we don’t need to expand the definition.

    for element in properties:
        value = properties[element]
        if SYM.LEFT_ROUND in value:
            tree[block_name][element] = parseFunc(value)
Enter fullscreen mode Exit fullscreen mode

Next we say if there is a ( in the value, it denotes a function like rgb(). In that case we use our function parseFunc.

        if SYM.DOT in value:
            tree[block_name][element] = memory[value.strip(SYM.DOT)]
Enter fullscreen mode Exit fullscreen mode

Next we see a . in the value, we go see in the memory dictionary and replace it

        if SYM.AT in element:
            del tree[block_name][element]
            tree[block_name].update(tree[element])
Enter fullscreen mode Exit fullscreen mode

Next we see the @ symbol as key in a block, we just add the dictionaries in it to the dictionaries in that block (Done by update() in Python) .

That also shows the advantage of using a data structure over plain strings.

Compiling To Normal Css

Now we can compile it to normal CSS, we’ll just print it for now (you can file= in print btw)

for key in tree:
    if key[0] == SYM.AT:
        continue
    print(key, '{', sep='')
    for elem in tree[key]:
        print('{}{}: {};'.format(SYM.TAB, elem, tree[key][elem]))
    print('}\n')
Enter fullscreen mode Exit fullscreen mode

which produces

.add{
    color: white;
    background-color: rgb(3,4,5);
}

#ad{
    color: rgb(1,4,5);
}

.a div a{
    color: berry;
    border: 10px solid black;
    border-radius: 50%;
}
Enter fullscreen mode Exit fullscreen mode

The Future of It

This was done without a safety net and lacking lot of cool features but it is done to demonstrate that with some normal programming logic, you can set up something cool. It was my usual linkedIn Python Project of The Week.

If you want to build a CSS pre-processor from scratch based on indentation (like stylus), use this tutorial.

What's the next thing/s to add? Comment it below!

Github Link: https://github.com/Abdur-rahmaanJ/dotdot

Mirrored on PythonMembersClub

Images from unsplash

— Abdur-Rahmaan Janhangeer

Top comments (2)

Collapse
 
abdurrahmaanj profile image
Abdur-Rahmaan Janhangeer •

Was also a featured article in here!
twitter.com/TechRunEn/status/11534...

Collapse
 
abdurrahmaanj profile image
Abdur-Rahmaan Janhangeer • • Edited

Featured on Awesome Python

python.libhunt.com/newsletter/165

Read next

aniruddhaadak profile image

🚀 Lucide React: High-Quality, Open-Source Icons for Your React Projects

ANIRUDDHA ADAK -

wearethreebears profile image

Exploring Typesafe design tokens in Tailwind 4

Liam Hall -

shittu_olumide_ profile image

How to Use PySpark for Machine Learning

Shittu Olumide -

habeebrahmanca07 profile image

Top 10 CSS Frameworks for Front-End Developers in 2024

Habeeb Rahman CA -