genro-builders

Builder system for genro-bag — domain-specific grammars, rendering, and runtime data binding via pointers, built on top of bag data structures.

Installation

pip install genro-builders

Quick start

A page is a builder: subclass the dialect and implement main.

from genro_builders.contrib.html import HtmlBuilder


class HelloPage(HtmlBuilder):
    def main(self, root):
        body = root.body()
        body.h1("Hello World")
        body.p("My first page with genro-builders.")


page = HelloPage()
page.create()
print(page.render(pretty=True))

The lifecycle is two-phase, both on the builder:

• create() runs setup(self.data) (seed the data), then the user-defined main(self.source) that populates the source Bag through the dialect's grammar API, then the first calculation of the data-elements.
• render(mode=None, target=None, **opts) drives the universal walk on the source and produces the dialect's output. Default mode comes from the dialect; default target returns the string. It composes two steps you can also call on their own: materialize(mode) walks and keeps the result in materialized[mode], finalize delivers it.

Rendering does not validate. validate_source() reports the nodes whose minimum child cardinality is unmet — (fullpath, [missing tags]) per node, empty when complete — when the author asks for it.

Dialects (contrib)

The package ships with reference dialects, each as a <Dialect>Builder grammar:

• HTML5 — genro_builders.contrib.html.HtmlBuilder (HTML5 grammar with CSS kwargs)
• SVG — genro_builders.contrib.svg.SvgBuilder
• CSS — genro_builders.contrib.css.CssBuilder
• XSD — genro_builders.xml (codegen: an XSD schema becomes a <Dialect>Builder you commit and import)

Mixed-dialect documents are supported via sub-builders: a grammar element marked _meta['subbuilder'] switches the active dialect from that node down, and the render walk picks the right renderer per node. HTML hosts SVG with body.svg(...); SVG hosts HTML with svg.html(...), wrapped in <foreignObject> automatically.

from genro_builders.contrib.html import HtmlBuilder


class Badge(HtmlBuilder):
    def main(self, root):
        body = root.body()
        svg = body.svg(viewBox="0 0 200 80", width=200, height=80)
        svg.rect(x=0, y=0, width=200, height=80, fill="#2c3e50")


page = Badge()
page.create()
print(page.render())

Architecture (one-paragraph map)

A builder declares the grammar of a dialect (decorators @element, @abstract; the three data-elements dataSetter / dataFormula / dataController are plain @element marked as data) and is also the document: it owns name, source, create()/render(), and exposes its renderers as renderer_<mode> properties. It also owns its datastore: one FLAT Bag, builder.data, with absolute paths and no leading segment — reachable from any node as node.data. A renderer is responsible for one mode: the universal walk produces fragments via dialect-specific rendered_item, then finalize ships the result to the target.

The builder also declares the class of its Source: _source_class (default SourceBag), redefinable per subclass. The Source declares the class of its nodes (_node_class). Branches follow the class of their parent bag. See the Source class.

Runtime data binding (pull-based)

Attribute values and node text can carry pointers and templates, resolved at render time:

• ^path — pointer declared as meant to follow the datum
• =path — passive pointer (read only)

Both resolve the same way in a static render: the difference is the author's declaration of intent, which a reactive engine would act on.

• ${name} — template token; an attribute referenced by a template of the same node is a consumed input, never emitted

from genro_builders.contrib.html import HtmlBuilder


class Page(HtmlBuilder):
    def setup(self, data):
        data.set_item("greeting", "Hello")

    def main(self, root):
        root.body().h1("^greeting")


page = Page()
page.create()                  # setup + main + the data-elements
print(page.render())
# ...<h1>Hello</h1>...

# Mutate the data and re-render: pull-based, no auto-render.
page.data.set_item("greeting", "Ciao")
print(page.render())
# ...<h1>Ciao</h1>...

Re-render is the whole reactivity model here: change the data, render again. Fine-grained reactivity is a separate engine, still under design — see the RX area of the contract.

The companion API on each source node:

• node.abs_datapath(path) — turn a relative path into an absolute one in the datastore
• node.get_relative_data(path) / node.set_relative_data(path, value) — read/write the datastore relative to the node
• node.SET / GET / PUT / FIRE — DSL macros over the same two entry points; the fired/reason flags they carry travel through the genro-bag subscribe pipeline, and nothing in this package acts on them (vocabulary for the future reactive engine)

Render target

page = HelloPage()
page.create()

# Return a string (default)
text = page.render()

# Write to a path
page.render(target="out.html")

# Push to a file-like or invoke a callable
import io
page.render(target=io.StringIO())
page.render(target=print)

# Register a default target (per mode), then render to it
page.set_render_target("out.html")
page.render()

Examples

Runnable tutorials under src/genro_builders/contrib/<dialect>/examples/, grouped by scenario:

• HTML — no_data/ (grammar, styling, sub-builders, validation, render modes), with_data/ (pointers, datapath, presentation), with_logic/ (data-elements)
• SVG — 01_introduction, badge_sheet, bar_chart, and more
• CSS — 01_introduction
• XSD — 01_person_schema; XSLT — 01_sitemap_to_html

Each example ships a runnable .py, a readme.md, and the rendered output. The test suite runs them all (tests/test_examples.py).

Documentation

• Getting Started — first page in 5 minutes
• Builders overview — builder/renderer split
• Decorators — @element, @abstract, sub-builders, data-elements
• Common patterns — ._ chaining, node_by_id, render targets
• Per-grammar references: HTML, SVG, CSS, XSD
• Architectural contract and roadmap: roadmap/

Downstream

genro-builders is a generic engine: the grammars, the renderers, and the pull-based data binding know nothing about who consumes them. The source carries no reference to any downstream layer — this is the one place that names them. Known consumers in the Genro ecosystem:

• genro-ws-web — WebSocket-driven reactive SPA framework (the production widget kit and the push transport live here)
• genro-office — Office document generation (Word and Excel builders)
• genro-print — print and PDF generation system
• genro-textual — Textual UI framework for Bag-driven applications
• genro-scriba — infrastructure configuration file generator (Traefik, Docker Compose, and more)

License

Apache License 2.0 — Copyright 2025 Softwell S.r.l.