Code Text Splitter
The Code Text Splitter is a specialized text splitter designed to split documents based on language-specific syntax. It utilizes the RecursiveCharacterTextSplitter from the LangChain library to perform intelligent splitting of code documents.
Node Details
-
Name: codeTextSplitter
-
Type: CodeTextSplitter
-
Version: 1.0
-
Category: Text Splitters
Parameters
1. Language
-
Type: Options
-
Description: The programming language of the code to be split.
-
Options:
-
cpp
-
go
-
java
-
js
-
php
-
proto
-
python
-
rst
-
ruby
-
rust
-
scala
-
swift
-
markdown
-
latex
-
html
-
sol
-
2. Chunk Size
-
Type: Number
-
Default: 1000
-
Optional: Yes
-
Description: The number of characters in each chunk. This determines the size of the text segments after splitting.
3. Chunk Overlap
-
Type: Number
-
Default: 200
-
Optional: Yes
-
Description: The number of characters to overlap between chunks. This helps maintain context between split segments.
Input/Output
Input
The node expects code or text input in the specified language.
Output
The node outputs split text chunks based on the specified parameters and language-specific syntax.
Usage
This node is particularly useful in workflows that involve processing or analyzing code, such as:
-
Code summarization
-
Code analysis tasks
-
Preparing code for language models
-
Splitting large codebases for easier processing
By respecting the syntax of the chosen programming language, it ensures that the splitting process maintains the logical structure of the code as much as possible, which can lead to better results in downstream tasks.
Implementation Details
The node uses the RecursiveCharacterTextSplitter.fromLanguage() method from LangChain, which applies language-specific splitting rules. This method is more intelligent than a simple character-based split, as it attempts to split at appropriate syntactic boundaries for the given language.
Note
The effectiveness of the splitting can vary depending on the complexity and structure of the input code. Users may need to experiment with different chunk sizes and overlaps to achieve optimal results for their specific use case.