Are you still struggling with reading such differencing results? Now try BDiff!:

BDiff is a text-based differencing algorithm that can identify accurate line-level and block-level differences between text files and generate corresponding edit scripts. It can be applied in scenarios such as code review and change analysis.

• Language-independent: BDiff is a text-based differencing algorithm and can thus be applied to code written in any programming language.
• Comprehensive Difference Identification: It can identify line deletions, line additions, line updates, line splits, line merges, block copies (including line updates), and block moves (including line updates).
• Generate Edit Scripts: Produces edit scripts based on the identified differences.
• Display Updated Difference Substrings: Shows the specific substrings that have been updated.
• Edit Action Localization: Helps pinpoint the edit actions within the script.
• View Before-and-After Mapped Lines/Blocks: Allows users to view the mapped lines or blocks before and after the changes.
• Set Difference Display Modes: Offers alignment mode and compact mode for displaying differences.
• Language Settings: Supports Chinese and English.
• Theme Settings: Switch between light and dark themes for comfortable viewing.
• Difference Option Settings: Customize comparison parameters for precise results.
• Programming Language Highlighting: Syntax-aware highlighting for major programming languages.

1. Visit the BDiff online tool at http://bdiff.net/.
2. Upload the old and new versions of your text file via the file selector.
3. Click to view the difference results.

• Git Difference Algorithm: Choose from Git's four differencing algorithms. Results may vary slightly; the default Histogram algorithm typically produces the shortest edit scripts.
• Tab Length: Number of spaces a tab character represents, affecting alignment calculations.
• Minimum Length of Moved Block: The minimum number of lines a moved block should contain.
• Minimum Length of Copied Block: The minimum number of lines a copied block should contain.
• Context Length: The number of lines above or below a line/block when calculating context similarity.
• Line Similarity Weight: The proportion of line similarity in the overall line mapping similarity calculation, which also takes into account context similarity.
• Overall Line Similarity Threshold: If the overall line mapping similarity is greater than or equal to this threshold, it is considered a valid line mapping.
• Maximum Merged Lines: The maximum number of lines allowed when detecting line merges.
• Maximum Split Lines: The maximum number of lines allowed when detecting line splits.
• Identify Block Moves: Enable/disable detection of block moves.
• Include Updates in Moves: Detect line updates within moved blocks.
• Include Lines with Only Stop Words in Moved Block Size Calculation: Whether to count lines containing only stop words when calculating the size of moved blocks.
• Maximal move-block distance: The maximum relative distance between the two blocks in a block-move edit action.
• Identify Block Copies: Enable/disable detection of block copies.
• Include Updates in Copies: Detect line updates within copied blocks.
• Include Lines with Only Stop Words in Copied Block Size Calculation: Whether to count lines containing only stop words when calculating the size of copied blocks.
• Maximal copying-block distance: The maximum relative distance between the two blocks in a block-copy edit action.

kitao/pyxel, 3861523a200da507f36edf478729f4ec7c269775, app.py

psf/requests, cde3b88f3e93a9503810acc0ded890025fcbc119, core.py

ansible/ansible, 3807824c6d0dae63b9f36dbafe8e100b0a3beaa6, init.py

topjohnwu/Magisk, fc5c9647d829cad1b73338e42164decc4ab08a54, drawer.xml

keras-team/keras, aa7f9cdae951bba824883cfa392224a292b284b, core.py

psf/black, e911c79809c4fd9b0773dea5b6a0e710b59614cf, test_black.py

wagtail/wagtail, a2a580f0fe7a1354a109eb062b5393fbb330f508, urls.py

square/okhttp, c8638813ff5f90715417e489b342aae5e410c5b2, pom.xml

scikit-learn/scikit-learn, 612312553118371289330f50b38653d1206246c0, gene.py

This software is licensed under Mulan Public License，Version 2 (Mulan PubL v2).

Copyright (c) [2025] [Lu YAO]
BDiff is licensed under Mulan PubL v2.
You can use this software according to the terms and conditions of the Mulan PubL v2.
You may obtain a copy of Mulan PubL v2 at:
         http://openworks.mulanos.cn/#/licenses/MulanPubL-v2
THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
See the Mulan PubL v2 for more details.

• Lu Yao
• Liu Wanwei
• Song Wansheng
• Chen Jing
• Yan Zhikang

We have conducted a comprehensive evaluation experiment on BDiff, with detailed results available in our paper: https://arxiv.org/abs/2510.21094. We are glad if you cite the relevant article in the following bibtex:

@misc{lu2025bdiffblockawareaccuratetextbased,
      title={BDiff: Block-aware and Accurate Text-based Code Differencing}, 
      author={Yao Lu and Wanwei Liu and Tanghaoran Zhang and Kang Yang and Yang Zhang and Wenyu Xu and Longfei Sun and Xinjun Mao and Shuzheng Gao and Michael R. Lyu},
      year={2025},
      eprint={2510.21094},
      archivePrefix={arXiv},
      primaryClass={cs.SE},
      url={https://arxiv.org/abs/2510.21094}, 
}