pyodt1

Python reimplementation of the Basic ODT (odt1) package.

Contents

• Getting started
  • What this project is
  • What you can do right now
  • Environment setup with uv
  • Run tests
  • Run the comparison harness
  • Build the docs locally
  • Suggested reading order for new learners
  • Important current scope caveat
• ODT for CFD readers
  • Why a CFD reader might find ODT strange at first
  • First principle: turbulent mixing is not only diffusion
  • Second principle: keep a real spatial line
  • Third principle: model stirring as discrete eddies
  • The triplet map: the heart of the stirring model
  • Why the middle copy is reversed
  • What an ODT realization looks like
  • Where randomness enters—and why
  • Candidate eddies versus accepted eddies
  • Why eddy size matters so much
  • Why acceptance probability matters
  • Deterministic advancement is still there
  • Why ODT can be powerful
  • What pyodt1 is doing specifically
  • How to connect the concepts to the package
  • A simple summary for CFD readers
  • What to keep in mind while learning ODT
• ODT in the literature
  • How to think about ODT in the wider turbulence-modeling landscape
  • A short reading path
  • Core references
  • 1. Classic formulation paper
  • 2. Vector formulation / free-shear-flow extension
  • 3. Early reacting-flow ODT application
  • 4. Broader conceptual overview
  • ODT beyond stand-alone 1D runs
  • 5. Near-wall LES closure based on ODT
  • 6. ODTLES report
  • More recent examples and practical reminders
  • 7. Modern stand-alone ODT for turbulent mixing
  • What these papers mean for pyodt1
  • Suggested use of this page
• Algorithm overview
  • Why ODT exists
  • High-level picture of one ODT step
  • The main conceptual ingredients
  • 1. A one-dimensional state
  • 2. Stochastic eddies
  • 3. Triplet map stirring
  • 4. Kernel-based velocity correction
  • How pyodt1 is organized
  • What is validated right now
  • What is not yet complete
  • How to read the rest of the docs
• Legacy Fortran to Python mapping
  • Upstream Fortran source
  • Current mapping
  • Notes
• Legacy RNG
  • Why match the original RNG?
  • Related Fortran files
  • Python implementation
  • The generator itself
  • State access
  • Validation status
  • Why this matters pedagogically
• Triplet map
  • Intuition
  • Discrete version in odt1
  • Python implementation
  • The add_k companion operation
  • Why the triplet map is useful in a reduced model
  • Implementation caution: indexing matters
  • Validation status
  • What new learners should take away
• Eddy sampling
  • Why sample L3 instead of L directly?
  • Related Fortran files
  • Step 1 — Build the length distribution
  • Step 2 — Sample L3
  • Step 3 — Compute L
  • Step 4 — Sample M
  • Why this stage matters physically
  • Why this stage matters computationally
  • Validation status
  • Recommended mental model for new learners
• Acceptance probability
  • Related Fortran file
  • Conceptual role
  • Inputs to the formula
  • Structure of the Fortran formula
  • Discrete correction factor
  • The role of PL
  • Important indexing note
  • Python implementation notes
  • Validation status
  • What new learners should remember
• Accepted eddy update
  • Related Fortran file
  • Overview
  • Step 1 — Kernel values
  • Step 2 — Coefficient construction
  • Step 3 — Triplet map
  • Step 4 — Add c*K
  • Why this update is interesting
  • Current validation status
  • Where this sits in the bigger algorithm
• Package structure
  • Roles
• Validation harness
  • Why a harness is necessary
  • Current comparison layers
  • 1. Fixed accepted-eddy path
  • 2. Sampled eddy path
  • 3. Deterministic advancement and initialization path
  • 4. BLowerdt and scheduled multi-trial realization behavior
  • 5. Repeated realizations and simplified statistics/output paths
  • 6. Eddy/change statistics and richer postprocessing
  • Usage
  • Behavior with and without Fortran
  • Why this matters for new contributors
• Fortran comparison status
  • Current status
  • Scope caveat
• Legacy source patches
  • Why patches were needed
  • Files patched
  • What was patched
  • 1. BRecord.f
  • 2. BSnap.f
  • What behavior changed
  • What behavior did not change
  • Validation status after patching
  • How to describe this in project status
  • Related pages
• Fortran routine status matrix
  • Status labels
  • Matrix
  • Important caveats
  • BInitIter.f
  • BSnap.f
  • Bodt.f
  • Recommended interpretation
• API
  • pyodt1 package

Project status

This project is an in-progress Python reimplementation of the Basic ODT (odt1) Fortran code.

Currently validated against the original Fortran implementation for a minimal path including:

• legacy RNG seeding and state progression

• eddy-size distribution setup

• eddy-size sampling

• eddy-location sampling

• BsKd

• acceptance probability

• accepted-eddy update (BEddy path)

Scope

The current focus is on:

1. making the core ODT algorithm readable in Python,

2. validating against the original Fortran code,

3. documenting algorithm and implementation details carefully.