numpy/matmul-diagonal
| Rule | matmul-diagonal |
|---|---|
| Pack | NumPy |
| Severity | warning |
| Fix | unsafe, with xpd check --fix --unsafe-fixes |
| Library versions | every version |
Computing a whole matrix product to read its diagonal or trace was 34 times slower than needed. Use np.einsum to compute only the diagonal.
leverage = np.diag(X @ cov @ X.T)leverage = np.einsum("ij,ij->i", X @ cov, X)What the rule reports
Section titled “What the rule reports”The rule reports np.diag(a @ b), np.diag(np.dot(a, b)), np.diagonal(a @ b), np.diag(a @ b.T), np.trace(a @ b) and np.trace(np.dot(a, b)).
How to fix it
Section titled “How to fix it”xpd check --fix --unsafe-fixes applies the fixes of this rule. Check the result, because an unsafe fix can change what the code does.
The unsafe fix writes np.einsum("ij,ji->i", a, b), np.einsum("ij,ij->i", a, b), np.einsum("ij,ji->", a, b).
Message
Section titled “Message”xpd prints this message with the issue. A name that starts with $ stands for code from your file.
This computes the whole product to read its diagonal, which ran 34 times slower on 1,000 by 1,000 matrices. Write $NP.einsum("ij,ji->i", $A, $B).Browse the other rules of the NumPy pack.