docs: content wave 1 (reference + how-to pages) - #449
Draft
lkdvos wants to merge 82 commits into
Draft
Conversation
Contributor
|
After the build completes, the updated documentation will be available here |
lkdvos
force-pushed
the
docs/content-wave1
branch
from
July 11, 2026 18:49
b4bf13e to
319a01f
Compare
Add the curated public-API contract page (grouped @ref links to the canonical docstrings) and expand the Library nav section in make.jl to hold the new per-category reference pages alongside the existing full index (lib/lib.md, now anchored as @id lib_index). Content-neutral: lib/lib.md stays the canonical @autodocs catch-all. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Category reference for the MPS state types via @docs; canonical=false (canonical home remains lib/lib.md), mirroring the man/algorithms.md convention. Quasiparticle types omitted pending docstrings (REVIEW flag). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Category reference for MPO and Hamiltonian types via @docs; canonical=false. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Category reference for the ground-state interface and algorithms via @docs; canonical=false. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a how-to guide with runnable recipes for constructing FiniteMPS, InfiniteMPS, WindowMPS, and MultilineMPS (random/initializer/element-type, product states, from tensors, multi-site unit cells, and U(1)-symmetric states), and introduce the How-to nav section in make.jl. All examples are @example blocks verified by the full build. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a how-to guide for inspecting, growing (RandExpand/OptimalExpand), and reducing (SvdCut) bond dimension, truncation schemes, dynamic growth via DMRG2/IDMRG2 and find_groundstate(...; trscheme), and algorithm chaining with &. All examples verified by the full build. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a how-to guide for local and multi-site expectation values (expectation_value), two-point correlators (correlator, incl. range form), full-MPO energies, and the energy variance convergence diagnostic. Wire the page into the How-to nav. All examples verified by the full build. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reflow prose to one sentence per line (semantic line breaks) and run the Runic formatter over the Julia code blocks across the Wave 1 reference and how-to pages. Whitespace/style only; no content or API changes, full build still green. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add 20 entries to the "Publications using MPSKit" list: 17 from the append-candidates review plus 3 found via web search (staelens2026, yang2026, zemlevskiy2026). Introduce a 2026 section and extend 2025. - Verified every arXiv ID / DOI resolves to a real paper. - Upgraded published preprints: vancraeynestdecuiper2026 to @Article (SciPost Phys. Lect. Notes 128); added arXiv eprints to the already published zong2026pseudogap, vandamme2025pseudogenerators, sommer2025. - Flagged with % NOTE the entries whose MPSKit citation could not be independently confirmed (lu2026 cites it only in its bibliography). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…volution, excitations) Category reference pages grouping the exported observables/analysis, bond-dimension, time-evolution, and excitation docstrings via `@docs; canonical=false`, mirroring the existing lib/ page structure. Every listed symbol verified exported with an attached docstring against src/. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two task-oriented how-to pages following the observables.md template: - howto/hamiltonians.md: finite/infinite MPO Hamiltonians from local terms, boundary-condition conversion, and window Hamiltonians. - howto/entanglement.md: entanglement entropy and spectrum from gauge tensors. All @example blocks verified to execute against real MPSKit; physics claims flagged with REVIEW for maintainer sign-off. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds howto/{hamiltonians,entanglement}.md to the How-to section and
lib/{bond_dimension,time_evolution,excitations,observables}.md to the Library section,
ordered to mirror the public API surface in lib/public.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Incorporate post-merge review feedback (docstring/wording only): - changebond docstring: drop inaccurate "in place", "enriched" -> "expanded" - OptimalExpand: reword "does not alter the state" -> "state-preserving" to avoid implying out-of-place operation - TDVP: "enriches" -> "expands" the bond Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Apply the new docstring style guide (docs/DOCSTRING_STYLE.md) across all of src/. Single-hash section headers, canonical section order and names (`Keyword Arguments`, not `Keywords`), dash bullets with types only where they earn their place, spaces around `=` in defaults and after commas in type-parameter lists. Templates applied: - A1 (`@kwdef` algorithm/config structs): $(TYPEDEF) + # Fields via $(TYPEDFIELDS) + # See also naming the consuming driver(s) + # References. - A2 (container/data types): $(TYPEDEF) + # Constructors + flat # Arguments/# Keyword Arguments + # Properties + # Notes. - B (user-facing verbs): stacked signatures with -> returns, # Arguments, # Keyword Arguments, # Returns. Also documents three previously-undocumented exports (find_groundstate!, leftenv, rightenv) and adds <!-- REVIEW --> flags for pre-existing doc inaccuracies surfaced during the sweep (non-existent Defaults.solver / Defaults.trscheme references, MPOHamiltonian constructor signatures with no matching methods) for maintainer follow-up. Verified: MPSKit precompiles and the full doctest suite passes. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address the batch of accurate/verifiable REVIEW comments left throughout the docs and package docstrings: - MPOHamiltonian: retarget the docstring constructor signatures and example to the concrete FiniteMPOHamiltonian / InfiniteMPOHamiltonian entry points (the MPOHamiltonian(lattice, ...) methods do not exist). - excitations umbrella docstring: drop the nonexistent `solver = Defaults.solver` default and the oversimplified FiniteExcited `init` default; describe the real forwarding/default behaviour. - ChepigaAnsatz2: document the real field defaults (Arnoldi(; krylovdim=30, tol=1e-10, eager=true) and notrunc()). - time_evolve / timestep: `alg` is a required positional argument, not defaulted to TDVP; correct both docstrings accordingly. - Fix a used-before-assignment bug in the InfiniteMPO QuasiparticleAnsatz excitations method (EffectiveExcitationHamiltonian(H_eff, ...) -> H). - observables reference page: surface the known source-status concerns for expectation_value(::MultilineMPS, ::MultilineMPO) and variance(::InfiniteQP, ::InfiniteMPOHamiltonian) as !!! warning admonitions. Doctests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Maintainer-confirmed resolutions for the remaining REVIEW comments: - hamiltonians.md: add a short TFIM description (X⊗X coupling, transverse field g along Z, quantum phase transition at g=1). - observables.md / entanglement.md: drop the connected-correlator, area-law, and sector-spectrum REVIEWs, leaving the text neutral (no phase-dependent physical interpretation asserted). - entanglement.md: replace the random-ψ∞ REVIEW with a !!! note clarifying it is not a converged ground state and pointing to find_groundstate. - entanglement.md: drop the WindowMPS REVIEW after verifying entropy(::WindowMPS, site) runs. - finite_envs.jl: drop the left/right environment REVIEWs; the wording matches the code (GRs[ind+1] contracts sites ind+1..end, gauge-compatible with the tensor at site ind, and mirror for leftenv). No REVIEW comments remain in src/ or docs/src/. Doctests pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…es, citation - CHANGELOG.md now canonical at repo root (visible on GitHub); make.jl copies it into docs/src at build time, generated copy gitignored - add CONTRIBUTING.md, issue templates (bug/question), PR template - CITATION.cff bumped to v0.13.13 (tag date 2026-06-09) - README gains a Citing section with BibTeX Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Replace the bare auto-generated title list with a curated gallery: difficulty ordering within each grouping, 1-2 sentence teasers stating the physics question and MPSKit features showcased, and thumbnail figures from the existing example outputs. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The trailing CLOSING NOTES maintainer-comment blocks claimed the pages' outputs were unverified and listed make.jl changes that have since landed. All @example blocks in the three pages were re-verified to execute end-to-end against the current package (32/32 pass, no code changes needed). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…tural) Maintainer-confirmed and cleared the final physics-judgment flags: - symmetry tutorial: g=2 symmetric GS, even/odd↔σˣ eigenstates, 8⊕8 block-sparsity speedup, paramagnetic single-spin-flip excitation, SU(2) multiplet gains - quench tutorial: tracking ⟨σˣ⟩ (⟨σᶻ⟩=0 by unbroken symmetry), relaxation with oscillations, post-quench entanglement growth capping the time window - structural: finite lazy vs infinite full gauge recompute; A=c·𝟙 → geometric/ power-law couplings; Vₗ⊗P←P⊗Vᵣ planarity/fermion rationale; "no algorithm wins everywhere" framing Only two non-physics notes remain (benchmark placeholder in index.md; a future-worked-example TODO in howto/excitations). Also restored a blank line before the gauge-freedom paragraph in matrix_product_states. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The home-page feature icons (finite-infinite, symmetry, algorithms, fast) 404'd: they were referenced by absolute URL /assets/icons/*.svg, but VitePress only serves absolute URLs from the public/ directory, and DocumenterVitepress auto-copies only top-level logo*/favicon* files from assets/ into public/ — so the icons/ subfolder was never published. Introduce the idiomatic VitePress public/ dir (Documenter mirrors the source tree, so docs/src/public/ lands at build/.documenter/public/, served at root). Move logo.svg, favicon.ico, and icons/*.svg there (creating src/public makes DocumenterVitepress skip its logo/favicon auto-copy, so those move too) and point the frontmatter icons at /icons/*.svg. logo_readme.svg stays in assets/ for the repo README. Verified with a full make.jl build: all six assets resolve at the site root and every index.html reference points to a real file. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…first - concepts/symmetries.md: flagship symmetry concept page (abelian / non-abelian / fermionic / anyonic, sectors, "when does SU(2) pay off"); every @example verified against real MPSKit; 4 REVIEW flags left for maintainer physics review. Anyonic section points at the hard-hexagon Fibonacci example. - examples gallery regrouped by computational task (groundstates / excitations / dynamics / statmech), replacing the quantum1d/classic2d split. Both make.jl files now auto-discover groups; Cache.toml re-keyed to new paths so the 10 existing examples are not rebuilt; embedded source paths + landing-page links updated. - new excitations/1.su3-heisenberg example reproducing the SU(3) [3 0 0] Heisenberg chain (devos2022) via SUNRepresentations.jl (added to the shared examples env). Hamiltonian built by hand (Casimir identity); computed [210] dispersion minimum is at k=2π/3 but ~-0.04, not the paper's +0.0263 — flagged, not hardcoded, pending maintainer physics review. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Completes the RESTRUCTURE_PLAN page tree. Each page was drafted against the real API (grepped/confirmed in src/), every @example executed against MPSKit, and the whole site verified with a full Documenter build (no broken @ref / checkdocs / duplicate-canonical / @example errors). Uncertain physics is marked with REVIEW comments for maintainer review. Concepts: - finite_vs_infinite.md — unit cells; per-site normalization (transfer-spectrum=1) - numerics.md — truncation (truncrank/trunctol/truncerror), convergence criteria, precision, pitfalls; defaults sourced from src/utility/defaults.jl How-to: - convergence_troubleshooting.md — symptom→diagnostic→fix recipes (variance check, changebonds, algorithm mixing, unit-cell/sector diagnostics) - statmech.md — transfer matrices, leading_boundary, MultilineMPS/MPO - quasi_1d_geometries.md — cylinders/ladders via MPSKitModels; ordering choices - saving_loading.md — stdlib Serialization + JLD2 round-trips (no bespoke MPSKit API) Library: - lib/environments.md, lib/internals.md — @docs canonical=false matching lib.md; internals marked non-public/unstable Meta: - migration.md, contributing.md, citing.md — grounded in CHANGELOG / CITATION.cff / Zenodo DOI / repo config (no invented facts) All new pages wired into docs/make.jl nav. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- accept textbook-fact claims (D³ cost, area law, finite-entanglement scaling, SU(2) multiplet-reduction payoff) and clear their flags; clear the remaining source-verified wording flags across numerics, convergence, finite_vs_infinite, symmetries, statmech, saving_loading - migration.md kept as a scaffold (changelog pointer); citing.md flags cleared - SU(3) example reframed off the gap: retitled "The SU(3) Heisenberg chain" and explains the dispersion is not a variational bound because the ground state is not converged tightly enough; regenerated via examples/make.jl - quasi_1d: document backandforth_pattern/frontandback_pattern as broken at MPSKitModels v0.4.7 (closure indexes a lazy iterator) plus the infinite-cylinder periodic-wrap caveat - keep the two forward-looking TODO comments (index benchmark, excitations example) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
After rebasing onto main (#452, sector-centered transfer spectrum): - transfer_spectrum keyword `num_vals` was renamed to `howmany` - correlation_length now returns a per-sector `SectorDict`; take the largest entry (`maximum(values(...))`) to recover the scalar correlation length the prose describes Fixes the two @example blocks (convergence_troubleshooting, statmech) that used the pre-#452 API. Full Documenter build is green. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Track the planning documents that guided the docs overhaul: RESTRUCTURE_PLAN.md (page tree + Diátaxis types, source of truth), IMPROVEMENT_PLAN.md (accessibility/adoption workstreams W0–W6), and REVIEW_SUMMARY.md (maintainer sign-off items from round 1). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Distilled review (2026-07) of quimb, TeNPy, and ITensor docs, with actionable items cross-referenced to RESTRUCTURE_PLAN.md. Companion to IMPROVEMENT_PLAN.md; salvaged from the docs-inspiration worktree before pruning it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add verbosity = 0 to the solver calls whose stderr logs leaked into four rendered example pages: the per-momentum "Found excitations" notices (hubbard, su3-heisenberg, haldane), the GradientGrassmann CG log (hubbard), and the DMRG2 groundstate sweep log (xy-finiteT). verbosity only controls logging, so computed results, plots and the intentional "┌ Info: Groundstate energy" summaries are unchanged. The obsolete log lines are stripped from the committed index.md artifacts (an emptied excitations output block in hubbard is removed) and the same edits are mirrored into the code-only main.ipynb notebooks, so the tree matches a fresh build. bose-hubbard keeps its 2-line VUMPS convergence summary (verbosity = 2) as an intentional healthy-convergence demonstration. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…oss-refs Add one-line descriptions to each lib/public.md category so it reads as a curated surface instead of a bare symbol list, add a short "How MPSKit compares" section to the README, and remove dead maintainer comments left over from the restructure (a resolved TODO(link) cross-reference, now wired via a real @id/@ref anchor on the examples gallery).
The example grew a bond of `FiniteMPS(ones(Float64, (ℂ^2)^4))` under `H = ΣXX` and asserted the bond dimension goes 1 -> 2. That state is |++++>, an exact eigenstate of every X⊗X term, so the two-site update projected onto the orthogonal complement is structurally zero (norm(g2) ~ 1e-16..1e-31). The old 1 -> 2 growth came from `normalize!` rescaling that noise to unit norm before the truncated SVD — precisely the pathology #464 fixed by skipping zero-content bonds, which turned this doctest red. Switch to `H = ΣZZ`, where |++++> has genuine expansion content (norm(g2) = 4.0 on every bond) and the bond really does grow 1 -> 2. Add a note documenting that bonds with no expansion content are left untouched, using the ΣXX case as the illustration. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The rebase onto main brought in docstrings written before the package-wide standardization, which deviated from docs/DOCSTRING_STYLE.md: - `AdaptiveKrylov`, `exact_diagonalization`, `MPSTensor`: `##`/`###` section headers and non-canonical spellings (`# Keyword arguments`) -> single-hash canonical names. - `DMRG3S` was the only driver-consumed `Algorithm` struct not on template A1: convert its hand-written signature to `$(TYPEDEF)` + `# Constructors`, add `# Fields`/`$(TYPEDFIELDS)` with per-field doc strings, and a `# See also` naming `DMRG`. - `LeftCanonical`/`RightCanonical`/`MixedCanonical`: add the missing `# See also` pointing at their driver `gaugefix!`. - Trailing `See also [...]` prose -> a `# See also` section. - `OperatorStyle`/`GeometryStyle`: `*` bullets -> `-`. - Spaces around `=` for defaults and after commas in type parameter lists, in bullet entries and in signature blocks alike (25 sites). Also adopt a blank line after every section header, package-wide (77 insertions across 30 files). The guide was silent on this and both styles were in use; document the rule in the quick rules and the section-header section, and fix templates A2 and B, whose snippets showed the other style. Finally, replace SketchedExpand's inline "(Gleis et al. Phys. Rev. Lett. 130, 246402 (2023))" with a `# References` section citing a new `gleis2023` bibliography entry, per the guide's rule that literature uses `@cite`. Entry verified against Crossref and arXiv:2207.14712. Full docs build green: doctests pass, no cross-reference or checkdocs errors. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Every other key in mpskit.bib is lowercase author+year; `Hubig2015` was the lone capitalized one. Rename it to `hubig2015` and update its single `@cite` in the DMRG3S docstring. Docs build green; the reference renders with the new anchor. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
lkdvos
force-pushed
the
docs/content-wave1
branch
from
July 27, 2026 23:08
62503ab to
2e45538
Compare
Both regressions come from how DocumenterVitepress emits docstrings, and neither is reachable from `make.jl`, so `src/.vitepress/config.mts` is now vendored from the DVP template with two marked additions. `make.jl` hashes the upstream template and warns when the copy needs re-syncing. #477: Documenter rewrites every heading inside a docstring into a bold-only paragraph (`recursive_heading_to_bold!`), so `# Constructors` arrives as `<p><strong>Constructors</strong></p>` with no `<h*>` to style. A markdown-it rule now tags exactly those paragraphs — scoped to docstring `<details>` blocks and requiring the paragraph to be nothing but one `<strong>`, so a paragraph merely starting with bold text is left alone — and `custom.css` gives them heading treatment that does not rely on font weight alone. The reason this only showed on Linux is a second bug: the DVP template heads `--vp-font-family-base` with "Barlow"/"Inter var" but never loads either as a webfont. macOS and Windows fall through to faces with real semibolds; Linux resolves a fontconfig substitute where weight 600 reads as regular. The stack is now upstream's minus the three families that are never available. #478: VitePress splits a page into search documents at headings only, and a docstring's anchor is an `<a id=...>` inside `<summary>`, so all 177 docstrings on `lib/lib` were indexed as a single document — no per-binding entry, poor ranking against short manual sections, and hits that could only link to the top of the page. A `search.options._render` hook rewrites each summary into a heading for indexing only. The injected markup has to put the title before the anchor: VitePress reads the section title from the text preceding it, and an empty title makes it drop the section silently. Verified on a full build: 298 headings tagged with no false positives and none outside docstrings; search `documentCount` 389 -> 647, exactly the 258 docstrings, with `lib/lib` split into 178 deep-linkable sections. Both fixes belong upstream in DocumenterVitepress; `UPSTREAM_DVP_NOTES.md` records what the upstream patches should look like so the vendored config can be dropped later. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Member
Author
|
@leburgel, any chance you can confirm this would resolve the two issues you brought up? |
Member
The search issue seems to be resolved, which was definitely the most important one. The docstring header issue improved a bit, but it's not quite solved. The custom |
Member
|
Unrelated: it seems the warning and note blocks have lost quite a bit of their style compared to the way they look in standard documenter. This is quite unfortunate, the blocks here look quite a bit worse than they used to. |
Follow-up to the #477/#478 fixes, addressing review feedback on #449. Bold still did not render bold. The remaining cause is VitePress' reset, which sets `font-synthesis: style` on <body>: listing only `style` permits synthetic oblique and *disables* synthetic bold, so wherever the resolved face has no real bold, `font-weight: 700` computes correctly yet still paints at regular weight. That is why the inspected weight looked right while the text did not. Allowing weight synthesis fixes it for docstring headings and for everything else subject to it — admonition titles, links inside callouts, docstring binding names and inline <strong>. Admonitions had lost their Documenter look for three compounding reasons: VitePress custom blocks carry no icon, no title colour and a transparent border; their title weight was subject to the same synthesis problem; and DocumenterVitepress maps every `!!! note` onto the `tip` container, whose dark-mode background its template overrides to a flat grey with dimmed text. Restyled towards Documenter: accent-coloured bold title with the same circle-exclamation icon, a 4px accent left rule, body text back at full contrast, and tip backgrounds restated so notes keep their colour in the dark theme. Everything is guarded by `:not(.jldocstring)`, since docstrings also carry the `custom-block` class. The note/tip distinction is deliberately not recovered. Both arrive as `::: tip` and the only signal left is the title, which Documenter defaults to "Note" — but 13 of this repo's 44 notes carry a custom title and are then indistinguishable from a real `!!! tip`. Colouring only the recognisable ones would be visibly inconsistent, so this waits on the upstream fix: VitePress 1.x has native `note`/`info` containers, making the remap obsolete. UPSTREAM_DVP_NOTES.md records both findings alongside the existing two. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two corrections to UPSTREAM_DVP_NOTES.md after checking the upstream tracker. The claim that VitePress 1.x ships native `note`/`important`/`caution` containers is wrong. Its `containerPlugin` registers only `tip`, `info`, `warning`, `danger` and `details`; the `.custom-block.note` and friends in its stylesheet belong to GitHub-style `> [!NOTE]` alerts. Verified against `createMarkdownRenderer`: `::: note Hello` comes out as a literal paragraph. So "stop remapping note to tip" is not a workable upstream fix, and it is also why LuxDL#289 — which passes every Documenter category straight through — cannot work as written. A fix has to keep mapping onto a registered container while preserving the original category for CSS, which is what LuxDL#288 and LuxDL#95 actually ask for. Also records the existing upstream issues and PRs per area, so the search work in LuxDL#385/#386 (MiniSearch tuning, orthogonal to our section splitting) and LuxDL#384 (a much larger opt-in replacement) are not duplicated. The docstring heading and font-synthesis findings appear to be unreported. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This was referenced Jul 30, 2026
lkdvos
added a commit
that referenced
this pull request
Jul 30, 2026
`first(js) > i || @error "..."` only logged the problem and then let the function continue into a contraction that is not the requested correlator, so an unordered (or equal) site pair returned a meaningless number instead of failing. Make it an `ArgumentError`, and document the ordering requirement in the docstring. All three `correlator` signatures funnel their arguments through the range method, so the test covers each of them. Carved out of #449 to keep that PR documentation-only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
lkdvos
added a commit
that referenced
this pull request
Jul 30, 2026
`excitations(H::InfiniteMPO, ::QuasiparticleAnsatz, ϕ₀::InfiniteQP, lenvs, renvs)` constructed its effective Hamiltonian as `EffectiveExcitationHamiltonian(H_eff, lenvs, renvs, E)`. `H_eff` is the local being assigned on that very line, so the read hit an uninitialized slot and every call through this method threw `UndefVarError: H_eff not defined in local scope` — the path was dead rather than silently wrong. The momentum entry point is unaffected, since for an `MPO` it converts to a `MultilineMPO` and dispatches to the `MultilineQP` method. The regression test uses the classical Ising transfer matrix rather than the six-vertex model of the neighbouring testset: with the latter the quasiparticle eigenproblem is only loosely converged (the eigenpair residual is ~1e-2 on both entry points), while for the former the two entry points agree to ~1e-11. Carved out of #449 to keep that PR documentation-only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
lkdvos
added a commit
that referenced
this pull request
Jul 30, 2026
`Base.isfinite(O::AbstractMPO) = isfinite(typeof(O))` requires every concrete operator type to define the trait, and `WindowMPOHamiltonian` did not, so both `isfinite(H)` and `isfinite(typeof(H))` threw a `MethodError`. A window Hamiltonian acts on a finite number of sites, so the trait is `true`, matching how it otherwise behaves as a finite Hamiltonian; it is spelled and placed like the `FiniteMPO(Hamiltonian)` definitions. Carved out of #449 to keep that PR documentation-only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
lkdvos
added a commit
that referenced
this pull request
Jul 31, 2026
`first(js) > i || @error "..."` only logged the problem and then let the function continue into a contraction that is not the requested correlator, so an unordered (or equal) site pair returned a meaningless number instead of failing. Make it an `ArgumentError`, and document the ordering requirement in the docstring. All three `correlator` signatures funnel their arguments through the range method, so the test covers each of them. Carved out of #449 to keep that PR documentation-only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
lkdvos
added a commit
that referenced
this pull request
Jul 31, 2026
`excitations(H::InfiniteMPO, ::QuasiparticleAnsatz, ϕ₀::InfiniteQP, lenvs, renvs)` constructed its effective Hamiltonian as `EffectiveExcitationHamiltonian(H_eff, lenvs, renvs, E)`. `H_eff` is the local being assigned on that very line, so the read hit an uninitialized slot and every call through this method threw `UndefVarError: H_eff not defined in local scope` — the path was dead rather than silently wrong. The momentum entry point is unaffected, since for an `MPO` it converts to a `MultilineMPO` and dispatches to the `MultilineQP` method. The regression test uses the classical Ising transfer matrix rather than the six-vertex model of the neighbouring testset: with the latter the quasiparticle eigenproblem is only loosely converged (the eigenpair residual is ~1e-2 on both entry points), while for the former the two entry points agree to ~1e-11. Carved out of #449 to keep that PR documentation-only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
lkdvos
added a commit
that referenced
this pull request
Jul 31, 2026
`Base.isfinite(O::AbstractMPO) = isfinite(typeof(O))` requires every concrete operator type to define the trait, and `WindowMPOHamiltonian` did not, so both `isfinite(H)` and `isfinite(typeof(H))` threw a `MethodError`. A window Hamiltonian acts on a finite number of sites, so the trait is `true`, matching how it otherwise behaves as a finite Hamiltonian; it is spelled and placed like the `FiniteMPO(Hamiltonian)` definitions. Carved out of #449 to keep that PR documentation-only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
lkdvos
added a commit
that referenced
this pull request
Jul 31, 2026
* fix: `correlator` throws on unordered sites `first(js) > i || @error "..."` only logged the problem and then let the function continue into a contraction that is not the requested correlator, so an unordered (or equal) site pair returned a meaningless number instead of failing. Make it an `ArgumentError`, and document the ordering requirement in the docstring. All three `correlator` signatures funnel their arguments through the range method, so the test covers each of them. Carved out of #449 to keep that PR documentation-only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: `excitations` on an `InfiniteMPO` with a quasiparticle input `excitations(H::InfiniteMPO, ::QuasiparticleAnsatz, ϕ₀::InfiniteQP, lenvs, renvs)` constructed its effective Hamiltonian as `EffectiveExcitationHamiltonian(H_eff, lenvs, renvs, E)`. `H_eff` is the local being assigned on that very line, so the read hit an uninitialized slot and every call through this method threw `UndefVarError: H_eff not defined in local scope` — the path was dead rather than silently wrong. The momentum entry point is unaffected, since for an `MPO` it converts to a `MultilineMPO` and dispatches to the `MultilineQP` method. The regression test uses the classical Ising transfer matrix rather than the six-vertex model of the neighbouring testset: with the latter the quasiparticle eigenproblem is only loosely converged (the eigenpair residual is ~1e-2 on both entry points), while for the former the two entry points agree to ~1e-11. Carved out of #449 to keep that PR documentation-only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * fix: `isfinite` for `WindowMPOHamiltonian` `Base.isfinite(O::AbstractMPO) = isfinite(typeof(O))` requires every concrete operator type to define the trait, and `WindowMPOHamiltonian` did not, so both `isfinite(H)` and `isfinite(typeof(H))` threw a `MethodError`. A window Hamiltonian acts on a finite number of sites, so the trait is `true`, matching how it otherwise behaves as a finite Hamiltonian; it is spelled and placed like the `FiniteMPO(Hamiltonian)` definitions. Carved out of #449 to keep that PR documentation-only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * shrink changelog --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Wave 1 of the documentation restructure (per
docs/RESTRUCTURE_PLAN.md).Opening as a draft so the Documenter/Vitepress PR preview build runs and the new pages can be reviewed online. Once CI finishes, the preview is deployed to:
https://quantumkithub.github.io/MPSKit.jl/previews/PR449/What's included
@autodocs)Notes
@example/jldoctestand were verified against real MPSKit.<!-- REVIEW: ... -->.🤖 Generated with Claude Code