Python API index
Every stable name Python exports, and its extension seams.
Package: earthsci_astdiff 0.2.0
Documentation extracted with griffe 2.3.2.
Extension-tier names are on the Python extension page .
Stable names
| Python | Canonical | Kind | Summary |
|---|---|---|---|
Band | band | type | One structured block of ∂(rhs of the equation for u)/∂v — see the Julia reference docstring and esm-jacobian-spec.md §4. |
BandError | band_error | error | Raised when an expression has a form the band calculus cannot differentiate (for example a joined faq node or a boolean semiring). |
bind_plan | bind_plan | function | Check plan against the state Jacobian of obj (an EsmFile, a flattened system or a SysView) and precompute its gather. |
BlockDiagonalPlan | block_diagonal_plan | type | Per-cell block LU of a block-diagonal Jacobian: the block variables in Markowitz order, the block pattern, the 1-based ordering and the pattern of the LU factors including fill-in. |
column_lines | column_lines | function | The per-line structure of a state Jacobian whose couplings all stay within one line along index set axis, or None when one does not (structure.jl column_lines): every band… |
ColumnBlockPlan | column_block_plan | type | Line-by-line factorization info for a "column_block" Jacobian: every coupling stays within one line along axis, so the matrix is one (len(block_vars)·line_length)-square… |
cse_templates | cse_templates | function | Greedy largest-first extraction of repeated subtrees. |
DerivativeRuleError | derivative_rule_error | error | Raised when a scalar operator has no derivative rule (spec §3). |
detect_structure | detect_structure | function | Classify the band list: "block_diagonal", "banded", "general" (or "empty" / "other"), read off the column-index expressions without evaluating anything. |
dscalar | dscalar | function | Derivative of the scalar expression e w.r.t. |
evaluation_document | evaluation_document | function | The derived evaluation document of the first-order bands on axis wrt (spec §8, model "JacobianEval"): the source’s observed the coefficients read, every source derivative… |
extended_bands | extended_bands | function | The extended first-order system of axis wrt (spec §4.5, ExtendedBands): every state equation’s bands with respect to the axis’s targets and the active recurrences, and each… |
forcing_names | forcing_names | function | The forcing set of a system, sorted by name: the parameters whose cells are the columns of the wrt="forcing" axis (esm-jacobian-spec.md §4.3). |
hvp_document | hvp_document | function | The second-order product document of spec §9.4 for the axis pair wrt and stages stage pairs: its right-hand side leaves g = ∇_{a2} Σᵢ μᵢᵀ J(u) wᵢ (∇_{a2} (Σᵢ wᵢ μᵢ)ᵀ f_t… |
JacEntry | jac_entry | type | One block of the Jacobian: row variable u, column variable v, and the Band giving the rows it covers, the column-index expressions and the coefficient. |
JacEntry2 | jac_entry2 | type | One second-order band: ∂²(equation for u)/∂v ∂w. |
jacobian_bands | jacobian_bands | function | Differentiate a model’s right-hand side into a list of Jacobian bands. |
jacobian_bands2 | jacobian_bands2 | function | Second-order bands for the axis pair wrt = (a1, a2), a1 in ("states", "time") and a2 in ("states", "parameters", "forcing"): every first-order entry of the a1 axis… |
jacobian_block | jacobian_block | function | The §4 jacobians block for one model: entries (template-CSE’d by default) and the block-local expression_templates; on the state and time axes the structure classification… |
jacobian_columns | jacobian_columns | function | The column space of a "parameters" or "forcing" block: each column variable with its integer shape ([] for a scalar), sorted by name — the columns field of the emitted… |
jacobian_pattern | jacobian_pattern | function | The structural sparsity pattern as sorted (rowname, colname) pairs plus the entries — dependency-light (no scipy). |
jvp_document | jvp_document | function | The J v product document of spec §9.2 for one first-order axis. |
lu_fill | lu_fill | function | Symbolic (no-pivoting) LU: the pattern plus every fill-in entry. |
markowitz_ordering | markowitz_ordering | function | Greedy Markowitz ordering (1-based), first minimal cost wins. |
parse_jacobian_blocks | parse_jacobian_blocks | function | parse_jacobian_block over a model’s jacobians entry, which is either one block or a list of blocks with distinct wrt (spec §4); returns one (wrt, entries) per block, in… |
plan_factorization | plan_factorization | function | Symbolic factorization info for the detected structure. |
recurrences | recurrences | function | Every recurrence of the view, in equation order: each observed whose definition reads itself, with its frame and axis read off the definition. |
scatter_plan | scatter_plan | function | Build the scatter plan of a model’s state Jacobian for layout. |
ScatterPlan | scatter_plan | type | |
serialize_band | serialize_band | function | One entry of the jacobians block, as a JSON-ready dict (spec §4; a JacEntry2 adds col2 / col2_idx, §4.4). |
serialize_plan | serialize_plan | function | |
simplify_branches | simplify_branches | function | Literal-condition fold, equal-branch collapse, same-condition pruning. |
Site | site | type | A bare variable reference or one symbolic index(v, i…) cell. |
stable_json | stable_json | function | Deterministic JSON: sorted object keys at every level, floats in the reference’s layout (canonical_float) — emit.jl’s stable_json. |
SysView | sys_view | type | A model or flattened system as the band calculus sees it: its variables, equations and index sets, with subsystems already expanded. |
sysview | sysview | function | View a model of an EsmFile, or a FlattenedSystem, for differentiation. |
transfer_definition | transfer_definition | function | The transfer recurrence T of rec (spec §4.5): over the frame plus one output index j on the axis, T[o, k, j] = [k == j] + Σ_{self bands} coef · T[o, k − lag, j], so `T[o,… |
vjp_document | vjp_document | function | The Jᵀ λ product document of spec §9.3 for one first-order axis (model "VjpEval"), as jvp_document: io["lam"] maps each state to its adjoint input and io["vjp"] each… |