Removing remnant legacy methods

[apchytr]

User description

Context: As an exercise, this PR aims to remove many of the methods that are no longer used in the core functionality of Mr Mustard.

Description of the Change: Updated DM.physical_stellar_decomposition_mixed to make use of math.pinv and updated the tensorflow pinv to account for the complex case. Updated several tests that globally mutated settings to instead make use of the context manager. Obliterated random.py. Removed the following:

backend_manager.py

• atleast_1d, atleast_2d, atleast_3d, block_diag, boolean_mask, constraint_func, convolution, repeat, round, set_diag, squeeze, cholesky, Categorical, MultivariateNormalTriL, single_mode_to_multimode_vec, single_mode_to_multimode_mat, poisson, binomial_conditional_prob, convolve_probs_1d, convolve_probs

bargmann_utils.py

• norm_ket, trace_dm

fock_utils.py

• ket_to_dm, ket_to_probs, dm_to_probs, U_to_choi, number_means, number_variances, validate_contraction_indices, sample_homodyne

gaussian.py

• number_cov, log_negativity

gaussian_integrals.py

• reorder_abc

Benefits: A more lightweight Mr Mustard experience.

Possible Drawbacks: Many of these methods may still be useful for user to have access to outside of the core MM functionality. In this case, this PR can always serve as reference for these methods and we can reintroduce them on a "as needed" basis.

---

PR Type

Other

---

Description

• Remove legacy unused methods from backend managers
• Update tests to use settings context manager
• Fix TensorFlow pinv for complex matrices
• Replace deprecated atleast_1d/2d/3d with atleast_nd

---

Changes walkthrough 📝

	Relevant files
Tests	4 files test_opt_lab.py Replace global settings mutation with context manager        +365/-357 test_callbacks.py Use settings context manager in tests                                        +34/-34  test_lattice_functions.py Use settings context manager in tests                                        +8/-8      test_b_to_ps.py Use settings context manager in tests                                        +2/-3
Enhancement	8 files backend_manager.py Remove legacy unused backend methods                                          +17/-363 backend_jax.py Remove legacy JAX backend methods                                                +1/-109  backend_numpy.py Remove legacy NumPy backend methods                                            +1/-134  backend_tensorflow.py Remove legacy methods and fix complex pinv                              +11/-67  fock_utils.py Replace atleast_1d with atleast_nd                                              +2/-125  dm.py Use math.pinv instead of numpy.linalg.pinv                              +1/-1      utils.py Replace atleast_1d with atleast_nd                                              +1/-1      circuits.py Replace atleast_1d with atleast_nd                                              +1/-1
Additional files	7 files bargmann_utils.py +0/-23    gaussian.py +0/-44    gaussian_integrals.py +0/-32    test_backend_manager.py +0/-129  test_bargmann_utils.py +0/-18    test_fock_utils.py +0/-26    test_gaussian_integrals.py +0/-23

---

Need help?

Type /help how to ... in the comments thread for any questions about Qodo Merge usage.

Check out the documentation for more information.

[qodo-merge-pro]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 4 🔵🔵🔵🔵⚪

🏅 Score: 82

🧪 PR contains tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Complex Matrix The new TensorFlow pinv implementation for complex matrices uses a custom algorithm that may not be numerically stable for all cases. The implementation should be validated against edge cases like singular matrices, matrices with very small eigenvalues, and matrices with different condition numbers. # need to handle complex case on our own # https://stackoverflow.com/questions/60025950/tensorflow-pseudo-inverse-doesnt-work-for-complex-matrices real_matrix = tf.math.real(matrix) imag_matrix = tf.math.imag(matrix) r0 = tf.linalg.pinv(real_matrix) @ imag_matrix y11 = tf.linalg.pinv(imag_matrix @ r0 + real_matrix) y10 = -r0 @ y11 return tf.cast(tf.complex(y11, y10), dtype=matrix.dtype) Missing Import The removal of scipy.special.binom import may cause issues if any remaining code still depends on binomial coefficient calculations. Need to verify all usages have been properly removed or replaced. from jax.errors import TracerArrayConversionError import numpy as np from scipy.stats import ortho_group, unitary_group from ..utils.settings import settings Numerical Stability The change from numpy.linalg.pinv to math.pinv may affect numerical stability and performance. The math.pinv implementation should handle edge cases consistently with the previous numpy implementation. gamma = math.pinv(R_c) @ R gamma_transpose = math.einsum("...ij->...ji", gamma)

[qodo-merge-pro]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
Possible issue	Add error handling for complex matrices The complex pseudo-inverse implementation may fail for singular matrices or when the intermediate matrix imag_matrix @ r0 + real_matrix becomes singular. Add error handling or fallback to the standard tf.linalg.pinv for real matrices to prevent runtime failures. mrmustard/math/backend_tensorflow.py [277-285] def pinv(matrix: tf.Tensor) -> tf.Tensor: - # need to handle complex case on our own - # https://stackoverflow.com/questions/60025950/tensorflow-pseudo-inverse-doesnt-work-for-complex-matrices + # Handle real matrices with standard pinv + if not tf.dtypes.as_dtype(matrix.dtype).is_complex: + return tf.linalg.pinv(matrix) + + # Complex case - custom implementation real_matrix = tf.math.real(matrix) imag_matrix = tf.math.imag(matrix) r0 = tf.linalg.pinv(real_matrix) @ imag_matrix - y11 = tf.linalg.pinv(imag_matrix @ r0 + real_matrix) + intermediate = imag_matrix @ r0 + real_matrix + y11 = tf.linalg.pinv(intermediate) y10 = -r0 @ y11 return tf.cast(tf.complex(y11, y10), dtype=matrix.dtype) Apply / Chat Suggestion importance[1-10]: 8 __ Why: The suggestion correctly points out that the custom pinv implementation is for complex matrices. Adding a check to use the standard tf.linalg.pinv for real matrices is a good practice that improves robustness and prevents potential runtime errors.	Medium
Update

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.28%. Comparing base (9d12319) to head (0883982).

Additional details and impacted files

@@             Coverage Diff             @@
##           develop     #608      +/-   ##
===========================================
+ Coverage    91.55%   92.28%   +0.72%     
===========================================
  Files          102      102              
  Lines         5828     5700     -128     
===========================================
- Hits          5336     5260      -76     
+ Misses         492      440      -52     

Files with missing lines	Coverage Δ
mrmustard/lab/circuits.py	91.93% <ø> (ø)
mrmustard/lab/states/dm.py	100.00% <100.00%> (ø)
mrmustard/lab/utils.py	95.65% <100.00%> (ø)
mrmustard/math/backend_manager.py	96.12% <100.00%> (+5.19%)	⬆️
mrmustard/physics/bargmann_utils.py	100.00% <ø> (ø)
mrmustard/physics/fock_utils.py	99.04% <100.00%> (+12.09%)	⬆️
mrmustard/physics/gaussian.py	97.50% <ø> (+18.65%)	⬆️
mrmustard/physics/gaussian_integrals.py	98.46% <ø> (+0.35%)	⬆️

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9d12319...0883982. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.