deterministic-docx-export 0.1.0

A deterministic library for exporting markdown content to Word (.docx) templates

Deterministic DOCX Export

A standalone Python library for exporting structured content (markdown or plain text) into Word (.docx) templates. This module is strictly non-generative - it does not call any LLMs and does not modify wording. It only maps existing content into Word document structures.

Author

Created and maintained by Ahsan Saeed.

This project is designed with a strong focus on determinism, correctness, and enterprise-grade document generation.

Features

• Scalar placeholder replacement: Replace simple placeholders like {{document_id}}, {{title}}, etc.
• Block placeholder replacement: Replace structured content blocks like {{summary}}, {{proposal}}, etc.
• Markdown parsing: Convert markdown content to Word structures (headings, lists, tables)
• Manual list rendering: Bullet and numbered lists use deterministic glyph/number insertion (no Word numbering XML) for stable, cross-platform rendering
• Plain text mode fallback: Paragraphs only, no structure inference
• Combined markdown export: Exports a combined markdown file (exported_markdown_content.md) for content analysis
• Flexible field mapping: Use any placeholder names via scalar_fields and block_fields dictionaries

Design Philosophy

This project prioritizes:

• Deterministic and reproducible output
• Exact content fidelity (no wording changes)
• Cross-platform Word rendering stability
• Explicit, manual control over list and layout behavior

Design decisions favor predictability and correctness over feature breadth.

List Rendering Approach

• Bullet lists: Manual glyph insertion with configurable glyph sets and indentation policies defined by ListRenderConfig. Maximum visual depth is configurable. Deep nesting behavior beyond max_visual_depth is deterministic and policy-driven (clamp_last, cycle, or textual strategies).
• Numbered lists: Manual hierarchical numbering (1., 1.1., 1.1.1.) with Python-based counter tracking per list block. Numbers reset for each new list block.
• Indentation: Both list types use configurable manual paragraph indentation (default: 0.25" per level, -0.25" hanging indent, configurable via ListRenderConfig)
• No Word numbering XML: Ensures stable rendering across platforms (especially macOS)

List Semantics

• Block-based rendering: List rendering is block-based, not AST-linked. Each markdown list block is processed independently and rendered as a separate visual block in Word.
• Visual preservation: Nested lists are preserved visually (indentation and glyphs reflect nesting depth), not structurally (no Word list object relationships).
• Mixed lists: Mixed bullet ↔ numbered lists are rendered as separate blocks by design. This is a deliberate stability decision to ensure deterministic output and cross-platform consistency. This is not a bug.

Deep Nesting

Very deep nesting (beyond max_visual_depth) may reduce readability due to Word layout constraints. This is a limitation of Word document layout, not the export engine. The engine handles deep nesting deterministically according to the configured strategy, but visual clarity may degrade with extreme nesting.

Compatibility

• Python: 3.8+
• Tested Platforms: macOS, Linux
• Windows: Supported (planned for full testing in future releases)

Dependencies

• pydantic>=2.0.0 - For data models
• python-docx>=1.1.0 - For Word document manipulation
• markdown-it-py>=3.0.0 - For markdown parsing

Installation

From PyPI

pip install deterministic-docx-export

From Source

git clone <repository-url>
cd deterministic-docx-export
pip install -r requirements.txt
pip install -e .

Testing

Run the test suite to verify functionality and inspect outputs:

# Install test dependencies
pip install -r requirements.txt

# Run all tests
pytest tests/

# Run with verbose output
pytest tests/ -v

# Run a specific test
pytest tests/test_word_export.py::TestBasicExport::test_basic_scalar_fields

Test outputs are saved in tests/test_output/ for inspection:

• Input JSON files (*_input.json) - The test data used
• Output DOCX files (*.docx) - The generated Word documents
• Markdown files (*_markdown.md) - Exported markdown content
• Config files (*_config.json) - Configuration used

See tests/README.md for more details.

Usage

Minimal Example

from pathlib import Path
from deterministic_docx_export.models.export_models import WordExportRequest
from deterministic_docx_export.services.word_export_service import export_to_word

# Create export request
request = WordExportRequest(
    scalar_fields={
        "title": "My Document",
        "author": "John Doe",
    },
    block_fields={
        "content": "# Introduction\n\nThis is the content with **bold** text.",
    }
)

# Export to Word
result = export_to_word(
    template_path=Path("template.docx"),
    request=request,
    markdown_mode=True,
    output_path=Path("output.docx"),
)

print(f"Word file: {result['word_file_path']}")

Basic Example

from pathlib import Path
from deterministic_docx_export.models.export_models import WordExportRequest
from deterministic_docx_export.services.word_export_service import export_to_word

# Create export request with dynamic fields
request = WordExportRequest(
    scalar_fields={
        "document_id": "DOC-12345",
        "title": "My Document",
        "author": "John Doe",
        "date": "2024-01-15",
    },
    block_fields={
        "introduction": "# Introduction\n\nThis is the introduction content...",
        "body": "## Main Content\n\nThis is the main body with **bold** and *italic* text.",
        "conclusion": "## Conclusion\n\nFinal thoughts here...",
    }
)

# Export to Word
result = export_to_word(
    template_path=Path("template.docx"),
    request=request,
    markdown_mode=True,  # Set to False for plain text mode
    output_path=Path("output/exported_document.docx"),
)

print(f"Word file saved to: {result['word_file_path']}")
if result['markdown_files']:
    print(f"Markdown file saved to: {result['markdown_files'][0]}")

Configuration Example

from pathlib import Path
from deterministic_docx_export.models.export_models import WordExportRequest
from deterministic_docx_export.models.export_config import ListRenderConfig
from deterministic_docx_export.services.word_export_service import export_to_word

# Create custom list rendering configuration
config = ListRenderConfig(
    max_visual_depth=7,  # Allow up to 7 nesting levels
    indent_inches_per_level=0.25,
    hanging_indent_inches=0.25,
    bullet_glyphs=("•", "◦", "▪", "▫", "▸", "▴", "▾"),
    deep_bullet_strategy="clamp_last",  # or "cycle" or "textual"
)

request = WordExportRequest(
    scalar_fields={"title": "Document with Deep Nesting"},
    block_fields={
        "content": """
# Deep Nested Lists

- Level 1
  - Level 2
    - Level 3
      - Level 4
        - Level 5
          - Level 6
            - Level 7
        """
    }
)

result = export_to_word(
    template_path=Path("template.docx"),
    request=request,
    markdown_mode=True,
    output_path=Path("output.docx"),
    config=config,
)

JSON Input Example

import json
from pathlib import Path
from deterministic_docx_export.models.export_models import WordExportRequest
from deterministic_docx_export.services.word_export_service import export_to_word

# Load from JSON
with open("data.json", "r") as f:
    data = json.load(f)

request = WordExportRequest(**data)

result = export_to_word(
    template_path=Path("template.docx"),
    request=request,
    markdown_mode=True,
    output_path=Path("output/exported_document.docx"),
)

Example data.json:

{
  "scalar_fields": {
    "document_id": "DOC-12345",
    "title": "My Document",
    "author": "John Doe"
  },
  "block_fields": {
    "introduction": "# Introduction\n\nContent here...",
    "body": "## Body\n\nMore content...",
    "conclusion": "## Conclusion\n\nFinal thoughts..."
  }
}

Template Placeholders

Scalar Placeholders

Scalar placeholders are replaced with simple text values. You can use any placeholder name you want by adding it to scalar_fields:

scalar_fields={
    "document_id": "DOC-123",
    "title": "My Document",
    "author": "John Doe",
    "custom_field": "Any value",
}

In your Word template, use {{document_id}}, {{title}}, {{author}}, {{custom_field}}, etc.

Block Placeholders

Block placeholders are replaced with structured content (markdown or plain text). You can use any placeholder name you want by adding it to block_fields:

block_fields={
    "introduction": "# Introduction\n\nContent...",
    "body": "## Body\n\nMore content...",
    "conclusion": "## Conclusion\n\nFinal thoughts...",
}

In your Word template, use {{introduction}}, {{body}}, {{conclusion}}, etc.

Markdown Support

When markdown_mode=True, the following markdown structures are supported:

• Headings: # H1, ## H2, etc. (up to H9)
• Paragraphs: Plain text paragraphs
• Bullet lists: - or * for unordered lists
• Numbered lists: 1. for ordered lists
• Tables: Markdown table syntax
• Inline formatting: **bold** and *italic* text (in headings and paragraphs, not in table cells)

Table Behavior

• Inline markdown formatting inside table cells is treated as literal text.
• Tables prioritize structure and layout determinism over inline formatting.

Output

The export function returns a dictionary with:

• word_file_path: Path to the generated DOCX file
• markdown_files: List containing the path to exported_markdown_content.md (empty if all blocks are empty)

Content Fidelity

This module respects content fidelity: all text is preserved exactly as provided, with only structural transformations applied (markdown syntax → Word objects).

Design Guarantees & Non-Goals

Guarantees

• Deterministic output: Given the same input, the export produces identical Word documents across runs and platforms.
• No Word numbering XML: Lists are rendered using manual glyph/number insertion, avoiding Word's numbering system for cross-platform stability.
• Cross-platform stability: Output renders consistently across Windows, macOS, and Linux Word viewers.
• Exact text fidelity: All text content is preserved exactly as provided, with no wording modifications.

Non-Goals

• Semantic AST reconstruction: The engine does not attempt to reconstruct semantic markdown AST relationships in Word. Lists are rendered visually, not structurally linked.
• Word auto-list compatibility: Lists are not converted to Word's native list objects. This is intentional for stability and determinism.
• Rich markdown inside tables: Inline markdown formatting (bold, italic) inside table cells is not processed. Table cells contain literal text only.

Versioning Strategy

This project follows Semantic Versioning (MAJOR.MINOR.PATCH):

• PATCH (0.1.x): Bug fixes, documentation improvements, internal optimizations
• MINOR (0.x.0): New features, enhancements, backward-compatible API additions
• MAJOR (x.0.0): Breaking changes, API modifications, changes to deterministic behavior

Note: v0.x indicates the API is still evolving. However, deterministic behavior guarantees will NOT change without a major version bump.

Stability Contract

• Deterministic output behavior is guaranteed within the same major version
• API changes within the same major version maintain backward compatibility
• Breaking changes require a major version increment

License

Licensed under the Apache License, Version 2.0. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

See LICENSE and NOTICE files for full license text and copyright information.

Publishing

For maintainers: To publish a new version to PyPI:

# Clean previous builds
rm -rf build dist *.egg-info

# Build distribution packages
python -m build

# Upload to PyPI (requires PyPI API token)
python -m twine upload dist/*

Note: Use PyPI API tokens for authentication. See PyPI documentation for token setup.