Odoo Abstract Model Generator from XSD Schemas

Generate Odoo abstract models (mixins) from XSD schemas using xsdata. Heavily inspired by xsdata-plantuml.

Video Tutorial: YouTube - xsdata-odoo Overview

---

Table of Contents

• Install
• Quick Start
• Real-World Usage
• Architecture Pattern
• Configuration
• Advanced Examples
• Field Prefixing Strategy
• Custom Filters
• Troubleshooting

---

Install

$ pip install xsdata[cli]
$ pip install git+https://github.com/akretion/xsdata-odoo

---

Quick Start

Microsoft Purchase Order Demo

$ xsdata generate tests/fixtures/po/po.xsd --output=odoo
Parsing schema po.xsd
Compiling schema po.xsd
Builder: 6 main and 1 inner classes
Analyzer input: 6 main and 1 inner classes
Analyzer output: 5 main and 1 inner classes
Generating package: generated.po

Brazilian NF-e (Electronic Invoice)

$ export XSDATA_SCHEMA=nfe
$ export XSDATA_VERSION=40
$ export XSDATA_SKIP="^ICMS.ICMS\d+|^ICMS.ICMSSN\d+"
$ export XSDATA_LANG="portuguese"

$ xsdata generate nfelib/nfe/schemas/v4_0 \
    --package nfelib.nfe.odoo.v4_0 \
    --output=odoo

---

Real-World Usage

xsdata-odoo is used in production by the OCA Brazilian Localization (l10n-brazil) to generate models for complex fiscal documents:

Brazilian Electronic Fiscal Documents

Document	Fields	Model	Version	OCA Module
NF-e (Nota Fiscal Eletrônica)	800+	55/65	4.0	l10n_br_nfe_spec
CT-e (Transport Document)	1000+	57	4.0	l10n_br_cte_spec
MDF-e (Manifest)	500+	58	3.0	l10n_br_mdfe_spec
SPED (Fiscal Reports)	Variable	-	-	l10n_br_sped_*

• NF-e: 800+ fields across multiple hierarchical structures (identification, items, taxes, transport, payment)
• CT-e: Multiple transport modes (road, air, waterway, rail, pipeline, multimodal) with specific fields each
• MDF-e: Aggregation of multiple NF-e/CT-e documents with municipal grouping
• SPED: Complex register hierarchies with parent-child relationships

Integration with nfelib: xsdata-odoo generates Odoo models, while nfelib handles XML serialization. Both use the same XSD schemas for consistency.

Architecture Pattern

Two-Layer Architecture

The OCA l10n-brazil uses a proven two-layer architecture:

1. Spec Modules (l10n_br_*_spec)

• 100% generated from XSD schemas
• Abstract models (mixins) with field definitions
• No business logic
• Versioned per schema version (nfe40, cte40, mdfe30)

2. Implementation Modules (l10n_br_nfe, etc.)

• Inherits from spec modules
• Maps to Odoo fiscal documents (l10n_br_fiscal.document)
• Business rules and validations
• Web service communication (SEFAZ)
• User interface

Benefits

• Schema Updates: Regenerate spec modules without touching business logic
• Version Migration: Different schema versions coexist (nfe40 vs nfe50)
• Testing: Business logic tested separately from generated code
• Maintainability: Clear separation of concerns

---

Configuration

Environment Variables

Variable	Description	Default
XSDATA_SCHEMA	Schema name	spec
XSDATA_VERSION	Schema version	10
XSDATA_SKIP	Regex patterns to skip (pipe-separated)	[]
XSDATA_LANG	Language for text processing	""
XSDATA_MONETARY_TYPE	XSD type to force fields.Monetary	""
XSDATA_NUM_TYPE	XSD type for numeric detection	TDec_[5:7.7:9]
XSDATA_CURRENCY_FIELD	Currency field name	currency_id
XSDATA_GENDS	GenerateDS compatibility mode	false

Python API

from xsdata_odoo import get_config

config = get_config()
print(config.schema)              # "spec"
print(config.version)             # "10"
print(config.field_safe_prefix)   # "spec10_"
print(config.inherit_model)       # "spec.mixin.spec"

---

Advanced Examples

Brazilian NF-e with ICMS Handling

The NF-e schema defines ICMS taxes with multiple groups (ICMS00, ICMS10, ICMS40, etc.) that share field names. Skip them to avoid conflicts:

export XSDATA_SCHEMA=nfe
export XSDATA_VERSION=40
export XSDATA_SKIP="^ICMS.ICMS\d+|^ICMS.ICMSSN\d+"
export XSDATA_LANG="portuguese"
export XSDATA_CURRENCY_FIELD="brl_currency_id"

xsdata generate nfelib/nfe/schemas/v4_0 \
  --package nfelib.nfe.odoo.v4_0 \
  --output=odoo

# Move generated files to your module
mv nfelib/odoo/nfe/v4_0/* your_addon/models/v4_0/

CT-e (Transport Document)

export XSDATA_SCHEMA=cte
export XSDATA_VERSION=40
export XSDATA_SKIP="^ICMS\d+|^ICMSSN+|ICMSOutraUF|ICMSUFFim|INFESPECIE_TPESPECIE"
export XSDATA_LANG="portuguese"

xsdata generate nfelib/cte/schemas/v4_0 \
  --package nfelib.cte.odoo.v4_0 \
  --output=odoo

MDF-e (Manifest Document)

export XSDATA_SCHEMA=mdfe
export XSDATA_VERSION=30
export XSDATA_LANG="portuguese"

xsdata generate nfelib/mdfe/schemas/v3_0 \
  --package nfelib.mdfe.odoo.v3_0 \
  --output=odoo

SPED Fiscal Reports (Advanced)

For complex non-XML fiscal reports with register hierarchies:

from xsdata_odoo.generator import OdooGenerator
from xsdata_odoo.filters import OdooFilters

class SpedFilters(OdooFilters):
    def registry_name(self, name="", parents=[], type_names=[]):
        # Custom naming: schema.version.register_code
        name = self.class_name(name)
        return f"{self.schema}.{self.version}.{name[-4:].lower()}"

    def class_properties(self, obj, parents):
        # Add custom metadata
        register = lookup_register(obj.name)
        return f"_sped_level = {register['level']}"

---

Field Prefixing Strategy

Problem

With 800+ fields in NF-e alone, plus thousands of OCA modules, field name collisions are a real risk. Additionally, schemas evolve (3.0 → 4.0 → 5.0).

Solution

Each field gets a prefix combining schema name + version digits:

• NF-e v4.0: nfe40_ prefix → nfe40_vBC, nfe40_vICMS
• CT-e v4.0: cte40_ prefix → cte40_vTPrest
• MDF-e v3.0: mdfe30_ prefix → mdfe30_qNFe

Versioning Strategy

Scenario	Example	Database Impact
Minor update (4.00 → 4.01)	Same nfe40_ prefix	Fields updated in place, --update sufficient
Major update (3.0 → 4.0)	nfe30_ → nfe40_ prefix	New fields/tables, data migration needed

Automatic Prefix Generation

from xsdata_odoo import get_config

config = get_config()
config.schema = "nfe"
config.version = "40"

print(config.field_safe_prefix)  # "nfe40_"

---

Custom Filters

Extend OdooFilters for specialized use cases:

from xsdata_odoo.filters import OdooFilters
from collections import OrderedDict

class MyCustomFilters(OdooFilters):
    def registry_name(self, name="", parents=[], type_names=[]):
        # Custom model naming logic
        return f"my_module.{self.version}.{name.lower()}"

    def _extract_field_attributes(self, parents, attr):
        # Add custom field attributes
        kwargs = super()._extract_field_attributes(parents, attr)

        # Add custom metadata
        kwargs["my_custom_attr"] = True

        return kwargs

# Use custom filters
generator = OdooGenerator(config)
generator.filters = MyCustomFilters(
    config,
    all_simple_types=[],
    all_complex_types=[],
    registry_names={},
    implicit_many2ones={},
)
generator.filters.register(generator.env)

---

Field Name Conflicts

Problem: XSD defines multiple groups with same field names (e.g., ICMS00.vBC, ICMS10.vBC).

Solution: Use XSDATA_SKIP to skip conflicting classes:

export XSDATA_SKIP="^ICMS.ICMS\d+|^ICMS.ICMSSN\d+"

Then implement business logic in your implementation module to handle the different ICMS groups.

Currency Fields

Problem: Monetary fields reference non-existent currency field.

Solution: Set XSDATA_CURRENCY_FIELD to match your Odoo module's currency field:

export XSDATA_CURRENCY_FIELD="company_currency_id"  # or "currency_id", "brl_currency_id"

Language-Specific Text Processing

Problem: Field labels not extracted properly from XSD documentation.

Solution: Set XSDATA_LANG for stopword processing:

export XSDATA_LANG="portuguese"  # or "english", "spanish", etc.

---

Links

• GitHub - xsdata-odoo
• GitHub - nfelib
• OCA l10n-brazil
• xsdata Documentation

License

MIT License - See LICENSE file for details

Copyright

2025 Akretion - Raphaël Valyi raphael.valyi@akretion.com