split_tokens.py

import json
import re

## split the code into multiple tokens so that it is not too long

def count_tokens(code):
    # Count the number of tokens in the code
    #tokens = re.findall(r'\b\w+\b', code)
    return len(code)

def split_code(code, max_tokens=2048):
    # Split the code into smaller chunks based on classes, functions, and other logical units
    class_pattern = re.compile(r'(?:class|struct|interface)\s+\w+\s*[\(\{]')
    function_pattern = re.compile(r'(?:def|function|func|fn)\s+\w+\s*\([^)]*\)\s*\{')
    import_pattern = re.compile(r'(?:import|include|require)\s+')
    comment_pattern = re.compile(r'(?://|#).*')

    # Remove comments from the code
    code = comment_pattern.sub('', code)

    # Find the starting points of classes, functions, and other logical units
    class_starts = [match.start() for match in class_pattern.finditer(code)]
    function_starts = [match.start() for match in function_pattern.finditer(code)]
    import_starts = [match.start() for match in import_pattern.finditer(code)]
    logical_units = sorted(class_starts + function_starts + import_starts + [len(code)])

    chunks = []
    start_index = 0
    for end_index in logical_units:
        chunk = code[start_index:end_index].strip()
        if chunk:
            token_count = count_tokens(chunk)
            if token_count <= max_tokens:
                chunks.append(chunk)
                start_index = end_index
            else:
                # If the chunk is too long, split it into smaller chunks
                sub_chunks = split_chunk(chunk, max_tokens)
                chunks.extend(sub_chunks)
                start_index = end_index

    return chunks

def split_chunk(chunk, max_tokens):
    # Split a chunk into smaller sub-chunks based on the maximum token limit
    sub_chunks = []
    start_index = 0
    while start_index < len(chunk):
        end_index = start_index + max_tokens
        sub_chunk = chunk[start_index:end_index].strip()
        sub_chunks.append(sub_chunk)
        start_index = end_index
    return sub_chunks

def find_split_index(code, max_tokens):
    # Find the index to split the code chunk
    token_count = 0
    for i, char in enumerate(code):
        if char.isspace():
            token_count += 1
            if token_count > max_tokens:
                return i
    return len(code)

def process_json_file(input_file, output_file):
    # Read the input JSON file
    with open(input_file, 'r') as file:
        data = json.load(file)

    # Process each entry in the JSON data
    output_data = []
    for entry in data:
        code = entry['content']
        token_count = count_tokens(code)
        if code != "":
            if token_count > 2048:
                # Split the code into smaller chunks
                chunks = split_code(code)

                # Create separate entries for each chunk
                for chunk in chunks:
                    output_data.append(json.dumps({"text": chunk}))
                    if len(chunk)>2048:
                        print("Chunks len: ", len(chunk))
                        print(chunk)
                        break
            else:
                # Create a single entry for the code
                output_data.append(json.dumps({"text": code}))

    # Save the output data to a new JSON file without square brackets
    with open(output_file, 'w') as file:
        file.write('\n'.join(output_data))

    print(f"Processing completed. Results saved to '{output_file}'.")

# Specify the input and output file paths
input_file = '../zkml-dataset/dataset.json'
output_file = 'tokenized_code_data.json'

# Process the JSON file
process_json_file(input_file, output_file)