Project Architecture¶

TASSEL is organized as a single Gradle module whose source lives under the net.maizegenetics package. Functionality is exposed through composable plugins that can be driven either from the Swing GUI or from the command-line pipeline.

TASSEL 5 package architecture

Source layout¶

tassel/
├── build.gradle.kts          # Gradle build configuration
├── settings.gradle.kts       # Gradle settings (root project name)
├── gradlew / gradlew.bat     # Gradle wrapper
├── scripts/                  # start_tassel / run_pipeline launcher scripts
├── docs/                     # MkDocs documentation site (this site)
└── src/
    ├── main/java/net/maizegenetics/   # main source (Java + Kotlin)
    └── test/java/net/maizegenetics/   # unit & integration tests

Top-level packages¶

All production code lives under src/main/java/net/maizegenetics/:

Package	Responsibility
`plugindef`	The plugin framework — `Plugin`, `AbstractPlugin`, `PluginParameter`, `DataSet`, `Datum`, and code generators. The backbone of TASSEL's extensibility.
`pipeline`	The command-line pipeline (`TasselPipeline`) that parses `-fork`/`-input`/`-combine` directives and chains plugins together.
`tassel`	The Swing desktop application, including `TASSELMainApp` (the GUI entry point).
`analysis`	The bulk of TASSEL's analytical functionality, grouped into sub-packages (see below).
`dna`	Genotype/DNA data models — genotype tables, SNPs, positions, maps, tags, and their I/O.
`phenotype`	Phenotype data models (attributes, traits, phenotype tables).
`taxa`	Taxa lists, taxa metadata, distance matrices, and trees.
`stats`	Statistical machinery — linear models, PCA, and general statistics utilities.
`matrixalgebra`	Matrix abstractions with EJML and native BLAS (JNI) backends.
`gui`	Reusable Swing widgets and dialogs.
`chart` / `progress`	Charting components and progress reporting.
`prefs`	User preferences.
`util`	Shared utilities used across the codebase.

The `analysis` sub-packages¶

net.maizegenetics.analysis is where most user-facing capabilities live:

Sub-package	Contents
`association`	GLM, MLM, fast multithreaded association, EQTL.
`modelfitter`	Stepwise additive model fitting.
`distance`	Kinship and distance matrices (centered/normalized IBS, A-matrix, dominance).
`popgen`	Population-genetics analyses (LD, diversity).
`imputation`	FILLIN, FSFHap, and numerical imputation methods.
`numericaltransform`	Numerical genotype/phenotype transforms and imputation-by-mean/kNN.
`filter`	Site/taxa/trait filtering plugins.
`data`	Import/export, merge, separate, and other data-management plugins.
`tree`	Tree building and the Archaeopteryx viewer.
`clustering`	Clustering analyses.
`chart`	Result plots (Manhattan, QQ, LD, charts).
`gbs`, `gbs/v2`, `gbs/repgen`	Genotyping-by-sequencing pipelines.
`phg`, `rna`, `avro`, `gobii`, `monetdb`, `b4r`	Integrations and specialized workflows.

The plugin model¶

Nearly every operation a user can perform is implemented as a plugin that extends net.maizegenetics.plugindef.AbstractPlugin. Plugins:

Declare their inputs and configuration as self-describing PluginParameter fields. This single declaration drives both the GUI dialog and the command-line flags — there is no separate CLI parser per plugin.
Implement processData(DataSet input) to do their work, receiving and returning a DataSet.
Are chained together by the pipeline package (for CLI use) or invoked from the GUI in tassel.

The unit of data exchange between plugins is a DataSet, a collection of Datum objects that each wrap a typed payload (for example a GenotypeTable, a Phenotype, or a DistanceMatrix) plus a name and comment.

To write your own plugin, see Developing Plugins.

GUI vs. pipeline¶

The same plugins power two front-ends:

GUI — net.maizegenetics.tassel.TASSELMainApp is the desktop application entry point (also the mainClass for ./gradlew run). Plugin PluginParameters are rendered as dialog fields.
Pipeline (CLI) — net.maizegenetics.pipeline.TasselPipeline parses a command string, instantiates plugins non-interactively, wires their inputs and outputs via -fork/-input/-combine, and runs them.

Because both front-ends share the same plugin code, an analysis available in the GUI is generally available on the command line as well.