Wishlist for PINT development

[abhisrkckl]

PINT Development Wishlist

This is a summary of features/enhancements/bugfixes for PINT. Some of them have been requested by users and some of them were proposed by developers.

Please suggest additions to the list in the comments below. Please also remind me to remove items that have been implemented. If you are interested in any of the issues, please let us know in the comments and/or create a new PR for it.

Items with an active/work in progress pull request are marked with (*).

New features

Timing model and parameters

• Support P0 and P1 in place of F0 and F1 (also #1330)
• Keith et al 2013 DM noise model (DMMODEL)
• Main sequence binary model (MSS)
• Multiple planet system models (BT1P and BT2P)
• Hyperbolic encounter events (Jennings et al 2020)
• Time jumps applied to Site Arrival Times (SATJUMP) and Barycentric Arrival Times (BATJUMP?)
• More elegant implementation of system-dependent FD parameters (FDJUMP)
  • Hacky implementation in #1615
• Variable-index chromatic delay, red noise, and events
  • Partially implemented in in #1616
  • Partially WIP in #1777
  • Partially WIP in #1797
• Red noise models (PLRedNoise and PLDMNoise) should support corner frequency (TNREDCORNER) and fundamental frequency of the basis (TNREDFLOW) as parameters
  • WIP in #1762
• PLDMNoise should work properly with wideband data
• Band noise and system noise
• Use CamelCase uniformly in timing model parameter class names
• Setting MJDParameter from a time object
• Improve/simplify parameter API
• More rigorous validation of timing model components
  • Partially implemented in in #1702
• Improve naming conventions for maskParameters
• Use a faster algorithm to solve Kepler equation

TOAs

• Support using pulse numbers for a subset of the TOAs
• Simplify working with TOA flags
• Check TOA flags for IPTA compliance

Residuals and Fitting

• Faster implementation of ECORR for chi2 and fitting
• Test whitened residuals for Gaussianity and whiteness
• Implement fftfit in Python instead of relying on PRESTO
  • WIP in #777
• Implement DRACULA using PINT
• Full support for Bayesian Single Pulsar Noise and Timing Analysis
  • Partially implemented in #1407
  • Partially implemented in #1426
• Support for derivative-aware MCMC methods (Requires derivatives w.r.t. noise parameters)
  • Partially implemented in #1639
• Implement robust fitting methods (e.g. Huber regression)
• Easy way to extract correlation matrices
• Implement epoch averaging independent of ECORR
• CompositeMCMCFitter is broken

Clocks, observatories and solar system ephemerides

• Easy way to pre-download clock files, ephemerides etc.
• Support INPOP ephemerides

Command line and GUI

• general2-like command line tool to be used with unix pipes

Platform support

• Support for platforms without extended precision (float80 or float128)

API and functionality improvements

• Improve random models module
  • WIP in #1334

Performance

• Make pintk more responsive
• Improve performance of orbit interpolator (used for satellite observatories)
• Speed up tests
• Speed up TOA file reading
• Speed up timing model evaluation
  • Partially done in #1646
  • Partially done in #1748

Code quality and tests

• Add type hints
• Improve unit test coverage
• Audit and clean up skipped and xpassed tests
• Review test tolerances
• Find and remove unused modules and dead code
• Test that the noise models in PINT are fully compatible with ENTERPRISE
• Make sure that PINT updates don't break downstream codes (pint_pal and ENTERPRISE)

Documentation and Examples

• Standardize citations

[dlakaplan]

I don't think there are bugs in the ELL1 model as-implemented. We are currently checking #1573 to add new terms.

[abhisrkckl]

Removed that.