matterviz 0.4.0 → 0.4.2

package/dist/trajectory/helpers.js

@@ -1,16 +1,14 @@
 // Shared utilities for trajectory parsing
 import { ATOMIC_NUMBER_TO_SYMBOL } from '../composition/parse';
-import { ELEM_SYMBOLS } from '../labels';
+import { is_elem_symbol } from '../element/helpers';
 import * as math from '../math';
-const element_symbol_set = new Set(ELEM_SYMBOLS);
+import { make_site } from '../structure/site';
 const is_valid_row = (row) => {
     if (!(Array.isArray(row) || (ArrayBuffer.isView(row) && `length` in row)))
         return false;
     return math.is_finite_vec3_like(row);
 };
 const is_valid_vec3 = (coords) => Array.isArray(coords) && math.is_finite_vec3_like(coords);
-export const is_valid_element_symbol = (symbol) => element_symbol_set.has(symbol);
-export const coerce_element_symbol = (symbol) => is_valid_element_symbol(symbol) ? symbol : undefined;
 // Validate that data is a proper 3x3 matrix
 // Accepts both regular arrays and typed arrays (Float32Array, Float64Array, etc.)
 export function validate_3x3_matrix(data) {
@@ -24,11 +22,18 @@ export function validate_3x3_matrix(data) {
 }
 export const convert_atomic_numbers = (numbers) => numbers.map((num) => {
     const symbol = ATOMIC_NUMBER_TO_SYMBOL[num];
-    if (!symbol || !is_valid_element_symbol(symbol)) {
+    if (!symbol || !is_elem_symbol(symbol)) {
         throw new Error(`Unknown atomic number in trajectory data: ${num}`);
     }
     return symbol;
 });
+// Tally per-element atom counts for trajectory metadata.element_counts
+export const count_elements = (elements) => {
+    const counts = {};
+    for (const element of elements)
+        counts[element] = (counts[element] || 0) + 1;
+    return counts;
+};
 export const create_structure = (positions, elements, lattice_matrix, pbc, force_data) => {
     if (positions.length !== elements.length) {
         throw new Error(`create_structure requires matching positions and elements lengths, got positions=${positions.length}, elements=${elements.length}`);
@@ -42,13 +47,7 @@ export const create_structure = (positions, elements, lattice_matrix, pbc, force
         const abc = cart_to_frac ? cart_to_frac(xyz) : [0, 0, 0];
         const force = force_data?.[idx];
         const properties = is_valid_vec3(force) ? { force } : {};
-        return {
-            species: [{ element: elements[idx], occu: 1, oxidation_state: 0 }],
-            abc,
-            xyz,
-            label: `${elements[idx]}${idx + 1}`,
-            properties,
-        };
+        return make_site(elements[idx], abc, xyz, `${elements[idx]}${idx + 1}`, properties);
     });
     return lattice_matrix
         ? {
@@ -126,39 +125,71 @@ export const read_ndarray_from_view = (view, ref) => {
                 throw new Error(`Unsupported shape`);
             })();
 };
-// Unified frame counting for XYZ
-export function count_xyz_frames(data) {
-    if (!data || typeof data !== `string`)
-        return 0;
-    const lines = data.trim().split(/\r?\n/);
-    let frame_count = 0;
+// Copy listed fields from source to target when they hold numbers
+export const copy_numeric_fields = (target, source, fields) => {
+    for (const field of fields) {
+        if (field in source && typeof source[field] === `number`)
+            target[field] = source[field];
+    }
+};
+// Max and RMS of per-atom force magnitudes, or null when no forces present. Loop-based
+// rather than Math.max(...spread) to avoid call-stack overflow on very large frames.
+export function calc_force_stats(forces) {
+    if (forces.length === 0)
+        return null;
+    let force_max = -Infinity;
+    let sum_sq = 0;
+    for (const force of forces) {
+        const magnitude = Math.hypot(...force);
+        if (magnitude > force_max)
+            force_max = magnitude;
+        sum_sq += magnitude ** 2;
+    }
+    return { force_max, force_norm: Math.sqrt(sum_sq / forces.length) };
+}
+// Walk concatenated (ext)XYZ frames in `lines`, yielding each frame's atom-count line
+// True when a whitespace-split line looks like an XYZ atom line: a short
+// non-numeric element token followed by three numeric coordinates. Shared by
+// trajectory frame iteration and structure-format sniffing so both stay in sync.
+export const is_xyz_atom_line = (parts) => parts !== undefined &&
+    parts.length >= 4 &&
+    isNaN(Number(parts[0])) &&
+    parts[0].length <= 3 &&
+    parts.slice(1, 4).every((coord) => coord !== `` && !isNaN(Number(coord)));
+// index, parsed atom count, and comment line. A candidate frame is accepted only when its
+// first few atom lines look like "<element> <x> <y> <z>"; otherwise we advance one line and
+// rescan. That validation doubles as content sniffing so numeric-leading non-XYZ formats
+// (e.g. VASP XDATCAR) aren't misread as frames, and keeps count_xyz_frames consistent with
+// the actual parse/index walk (both go through this single source of truth).
+export function* iter_xyz_frames(lines) {
     let line_idx = 0;
     while (line_idx < lines.length) {
-        if (!lines[line_idx]?.trim()) {
-            line_idx++;
-            continue;
-        }
-        const num_atoms = parseInt(lines[line_idx].trim(), 10);
+        const num_atoms = Math.trunc(Number(lines[line_idx]?.trim()));
         if (isNaN(num_atoms) || num_atoms <= 0 || line_idx + num_atoms + 2 > lines.length) {
-            line_idx++;
+            line_idx++; // skip blank/invalid lines until the next frame's atom-count line
             continue;
         }
-        // Quick validation of first few atom lines
         let valid_coords = 0;
-        for (let idx = 0; idx < Math.min(num_atoms, 3); idx++) {
-            const parts = lines[line_idx + 2 + idx]?.trim().split(/\s+/);
-            if (parts?.length >= 4 && isNaN(parseInt(parts[0], 10)) && parts[0].length <= 3) {
-                if (parts.slice(1, 4).every((coord) => !isNaN(parseFloat(coord))))
-                    valid_coords++;
-            }
-        }
-        if (valid_coords >= Math.min(num_atoms, 3)) {
-            frame_count++;
-            line_idx += 2 + num_atoms;
+        const sample = Math.min(num_atoms, 3);
+        for (let idx = 0; idx < sample; idx++) {
+            if (is_xyz_atom_line(lines[line_idx + 2 + idx]?.trim().split(/\s+/)))
+                valid_coords++;
         }
-        else {
-            line_idx++;
+        if (valid_coords < sample) {
+            line_idx++; // count line looks valid but atom lines don't — likely non-XYZ content
+            continue;
         }
+        yield { start: line_idx, num_atoms, comment: lines[line_idx + 1] || `` };
+        line_idx += num_atoms + 2;
     }
+}
+// Count XYZ frames via iter_xyz_frames so total_frames matches what gets indexed/loaded
+export function count_xyz_frames(data) {
+    if (!data || typeof data !== `string`)
+        return 0;
+    const frames = iter_xyz_frames(data.trim().split(/\r?\n/));
+    let frame_count = 0;
+    while (!frames.next().done)
+        frame_count += 1;
     return frame_count;
 }

package/dist/trajectory/index.d.ts

@@ -7,6 +7,7 @@ export { default as TrajectoryExportPane } from './TrajectoryExportPane.svelte';
 export { default as TrajectoryInfoPane } from './TrajectoryInfoPane.svelte';
 export type TrajectoryFormat = `hdf5` | `json` | `xyz` | `xdatcar` | `traj` | `unknown`;
 export type { AtomTypeMapping } from './types';
+export declare const FRAME_LOAD_DEBOUNCE_MS = 75;
 export interface ParseProgress {
     current: number;
     total: number;

package/dist/trajectory/index.js

@@ -2,6 +2,9 @@ export { default as Trajectory } from './Trajectory.svelte';
 export { default as TrajectoryError } from './TrajectoryError.svelte';
 export { default as TrajectoryExportPane } from './TrajectoryExportPane.svelte';
 export { default as TrajectoryInfoPane } from './TrajectoryInfoPane.svelte';
+// Debounce for on-demand frame loads while scrubbing: skips fetches for steps
+// the user slides past. Exported so tests stay in sync with the real delay.
+export const FRAME_LOAD_DEBOUNCE_MS = 75;
 export function validate_trajectory(trajectory) {
     // with detailed error reporting
     const errors = [];

package/dist/trajectory/parse/ase.d.ts

@@ -1,2 +1,10 @@
-import type { TrajectoryType } from '../index';
+import type { TrajectoryFrame, TrajectoryType } from '../index';
+export declare const read_ase_header: (view: DataView) => {
+    n_items: number;
+    offsets_pos: number;
+};
+export declare function decode_ase_frame(view: DataView, buffer: ArrayBuffer, frame_offset: number, step: number, fallback_numbers?: number[], max_json_length?: number): {
+    frame: TrajectoryFrame;
+    numbers: number[];
+};
 export declare function parse_ase_trajectory(buffer: ArrayBuffer, filename?: string): TrajectoryType;

package/dist/trajectory/parse/ase.js

@@ -1,18 +1,55 @@
 // ASE trajectory (.traj) parsing - binary format
+import * as math from '../../math';
 import { MAX_SAFE_STRING_LENGTH } from '../constants';
 import { convert_atomic_numbers, create_trajectory_frame, read_ndarray_from_view, validate_3x3_matrix, } from '../helpers';
+// ULM header: frame count lives at byte 32, frame-offsets table position at byte 40
+export const read_ase_header = (view) => ({
+    n_items: Number(view.getBigInt64(32, true)),
+    offsets_pos: Number(view.getBigInt64(40, true)),
+});
+// Decode a single ASE/ULM frame (JSON header + optional ndarray payloads) into a
+// TrajectoryFrame. Returns the atomic numbers actually used so callers can cache them
+// as fallback for later frames that omit `numbers` (ASE stores them only once).
+export function decode_ase_frame(view, buffer, frame_offset, step, fallback_numbers, max_json_length) {
+    const json_length = Number(view.getBigInt64(frame_offset, true));
+    if (max_json_length !== undefined && json_length > max_json_length) {
+        throw new Error(`frame JSON too large: ${json_length} bytes`);
+    }
+    const frame_data = JSON.parse(new TextDecoder().decode(new Uint8Array(buffer, frame_offset + 8, json_length)));
+    const positions_ref = frame_data[`positions.`] ?? frame_data.positions;
+    const positions = positions_ref?.ndarray
+        ? read_ndarray_from_view(view, positions_ref)
+        : positions_ref;
+    const numbers_ref = frame_data[`numbers.`] ?? frame_data.numbers ?? fallback_numbers;
+    const numbers = numbers_ref?.ndarray
+        ? read_ndarray_from_view(view, numbers_ref).flat()
+        : numbers_ref;
+    if (!numbers || !positions) {
+        throw new Error(`missing ${!numbers ? `numbers` : `positions`}`);
+    }
+    const cell = frame_data.cell ? validate_3x3_matrix(frame_data.cell) : undefined;
+    const metadata = {
+        step,
+        ...frame_data.calculator,
+        ...frame_data.info,
+    };
+    if (cell) {
+        try {
+            metadata.volume = Math.abs(math.det_3x3(cell));
+        }
+        catch (error) {
+            console.warn(`Failed to calculate volume for frame ${step}:`, error);
+        }
+    }
+    const frame = create_trajectory_frame(positions, convert_atomic_numbers(numbers), cell, frame_data.pbc ?? [true, true, true], step, metadata);
+    return { frame, numbers };
+}
 export function parse_ase_trajectory(buffer, filename) {
     const view = new DataView(buffer);
-    let offset = 0;
     const signature = new TextDecoder().decode(new Uint8Array(buffer, 0, 8));
     if (signature !== `- of Ulm`)
         throw new Error(`Invalid ASE trajectory`);
-    offset += 24;
-    // Skip ASE/Ulm version field; current parsing logic is version-independent.
-    offset += 8;
-    const n_items = Number(view.getBigInt64(offset, true));
-    offset += 8;
-    const offsets_pos = Number(view.getBigInt64(offset, true));
+    const { n_items, offsets_pos } = read_ase_header(view);
     if (n_items <= 0)
         throw new Error(`Invalid frame count`);
     if (offsets_pos < 0 || offsets_pos + n_items * 8 > buffer.byteLength) {
@@ -23,31 +60,9 @@ export function parse_ase_trajectory(buffer, filename) {
     let global_numbers;
     for (let idx = 0; idx < n_items; idx++) {
         try {
-            offset = frame_offsets[idx];
-            const json_length = Number(view.getBigInt64(offset, true));
-            offset += 8;
-            if (json_length > MAX_SAFE_STRING_LENGTH) {
-                console.warn(`Skipping frame ${idx + 1}/${n_items}: too large`);
-                continue;
-            }
-            const frame_data = JSON.parse(new TextDecoder().decode(new Uint8Array(buffer, offset, json_length)));
-            const positions_ref = frame_data[`positions.`] ?? frame_data.positions;
-            const positions = positions_ref?.ndarray
-                ? read_ndarray_from_view(view, positions_ref)
-                : positions_ref;
-            const numbers_ref = frame_data[`numbers.`] ?? frame_data.numbers ?? global_numbers;
-            const numbers = numbers_ref?.ndarray
-                ? read_ndarray_from_view(view, numbers_ref).flat()
-                : numbers_ref;
-            if (numbers)
-                global_numbers = numbers;
-            if (!numbers || !positions) {
-                console.warn(`Skipping ASE frame ${idx + 1}/${n_items}: missing ${!numbers ? `numbers` : `positions`}`);
-                continue;
-            }
-            const elements = convert_atomic_numbers(numbers);
-            const metadata = { step: idx, ...frame_data.calculator, ...frame_data.info };
-            frames.push(create_trajectory_frame(positions, elements, frame_data.cell ? validate_3x3_matrix(frame_data.cell) : undefined, frame_data.pbc ?? [true, true, true], idx, metadata));
+            const { frame, numbers } = decode_ase_frame(view, buffer, frame_offsets[idx], idx, global_numbers, MAX_SAFE_STRING_LENGTH);
+            global_numbers = numbers;
+            frames.push(frame);
         }
         catch (error) {
             console.warn(`Error processing frame ${idx + 1}/${n_items}:`, error);

package/dist/trajectory/parse/diagnostics.d.ts

@@ -0,0 +1,3 @@
+export declare const get_traj_parse_warnings: () => string[];
+export declare const reset_traj_parse_warnings: () => void;
+export declare const traj_warn: (message: string, error?: unknown) => void;

package/dist/trajectory/parse/diagnostics.js

@@ -0,0 +1,14 @@
+import { to_error } from '../../utils';
+// Collector for non-fatal trajectory parse warnings (skipped atoms, dropped frames, plot-metadata extraction failures, ...) so they reach the UI not just the console. Reset at the start of each top-level parse call; format parsers append via traj_warn (mirrored to console.warn). Fatal failures throw — see parse_trajectory_data.
+let traj_parse_warnings = [];
+// Read-only snapshot of warnings since the last top-level parse (attached to trajectory metadata by parse/index.ts for UI surfacing)
+export const get_traj_parse_warnings = () => [...traj_parse_warnings];
+export const reset_traj_parse_warnings = () => void (traj_parse_warnings = []);
+export const traj_warn = (message, error) => {
+    const detail = error === undefined ? `` : `: ${to_error(error).message}`;
+    traj_parse_warnings.push(`${message}${detail}`);
+    if (error === undefined)
+        console.warn(message);
+    else
+        console.warn(`${message}:`, error);
+};

package/dist/trajectory/parse/h5-utils.d.ts

@@ -0,0 +1,13 @@
+import type { ElementSymbol } from '../../element/types';
+import type { Matrix3x3 } from '../../math';
+import type { Dataset, Entity, Group } from 'h5wasm';
+import * as h5wasm from 'h5wasm';
+export declare const is_hdf5_dataset: (entity: Entity | null) => entity is Dataset;
+export declare const is_hdf5_group: (entity: Entity | null) => entity is Group;
+export declare const read_dataset: (h5_file: h5wasm.File, path: string) => unknown;
+export declare const to_string_array: (data: unknown) => string[] | null;
+export declare const to_number_array: (data: unknown) => number[] | null;
+export declare const to_scalar_number: (data: unknown) => number | null;
+export declare const scale_matrix: (matrix: Matrix3x3, scale: number) => Matrix3x3;
+export declare const expand_ion_types: (ion_types: string[], ion_counts: number[]) => ElementSymbol[];
+export declare function with_h5_file<T>(buffer: ArrayBuffer, filename: string | undefined, callback: (h5_file: h5wasm.File) => T | Promise<T>): Promise<T>;

package/dist/trajectory/parse/h5-utils.js

@@ -0,0 +1,109 @@
+// Shared h5wasm plumbing for HDF5 trajectory parsers: type guards plus the
+// write-to-in-memory-FS/open/cleanup lifecycle both torch-sim and vaspout
+// parsing need (the FS write is the expensive part, so parsers share one open
+// file handle instead of re-opening per format probe).
+import { is_elem_symbol } from '../../element/helpers';
+import * as h5wasm from 'h5wasm';
+export const is_hdf5_dataset = (entity) => entity !== null && `to_array` in entity && entity instanceof h5wasm.Dataset;
+export const is_hdf5_group = (entity) => entity !== null && `keys` in entity && entity instanceof h5wasm.Group;
+// Datasets in interrupted files can be torn mid-chunk, making to_array throw —
+// treat unreadable datasets like missing ones so callers keep what parsed so far.
+export const read_dataset = (h5_file, path) => {
+    try {
+        const entity = h5_file.get(path);
+        return is_hdf5_dataset(entity) ? entity.to_array() : null;
+    }
+    catch {
+        return null;
+    }
+};
+// h5wasm returns JS strings for variable-length UTF-8 but may hand back byte
+// arrays for fixed-length string datasets (like VASP's |S2 ion_types).
+export const to_string_array = (data) => {
+    if (!Array.isArray(data))
+        return null;
+    const decoder = new TextDecoder();
+    const strings = [];
+    for (const item of data) {
+        if (typeof item === `string`)
+            strings.push(item.trim());
+        else if (item instanceof Uint8Array)
+            strings.push(decoder.decode(item).trim());
+        else
+            return null;
+    }
+    return strings;
+};
+// Integer datasets (like int64 number_ion_types) can surface as BigInt
+// values or typed arrays depending on dtype; normalize to plain numbers.
+export const to_number_array = (data) => {
+    const values = Array.isArray(data)
+        ? data
+        : ArrayBuffer.isView(data)
+            ? Array.from(data)
+            : null;
+    if (!values)
+        return null;
+    const numbers = values.map((item) => (typeof item === `bigint` ? Number(item) : item));
+    return numbers.every((item) => typeof item === `number` && Number.isFinite(item))
+        ? numbers
+        : null;
+};
+// Scalar datasets surface as plain numbers, 1-element arrays, or BigInt
+// depending on dtype; normalize to a finite number or null.
+export const to_scalar_number = (data) => {
+    const value = Array.isArray(data) ? data[0] : data;
+    if (typeof value === `bigint`)
+        return Number(value);
+    return typeof value === `number` && Number.isFinite(value) ? value : null;
+};
+// Apply VASP's POSCAR-style universal scaling factor to lattice vectors
+export const scale_matrix = (matrix, scale) => scale === 1 ? matrix : matrix.map((row) => row.map((val) => val * scale));
+// Expand VASP's (ion_types, number_ion_types) pair into per-atom symbols,
+// e.g. ([Ga, Sb], [1, 1]) -> [Ga, Sb]. Throws on unknown symbols and
+// malformed counts so torn/corrupt HDF5 input fails loudly, not by dropping atoms.
+export const expand_ion_types = (ion_types, ion_counts) => {
+    if (ion_types.length !== ion_counts.length) {
+        throw new Error(`ion_types (${ion_types.length}) and ion_counts (${ion_counts.length}) length mismatch`);
+    }
+    const elements = [];
+    for (const [type_idx, symbol] of ion_types.entries()) {
+        if (!is_elem_symbol(symbol)) {
+            throw new Error(`Unknown element symbol in ion_types: ${symbol}`);
+        }
+        const ion_count = ion_counts[type_idx];
+        if (!Number.isFinite(ion_count) || !Number.isInteger(ion_count) || ion_count < 0) {
+            throw new Error(`Invalid ion count for ${symbol}: ${ion_count}`);
+        }
+        for (let count = 0; count < ion_count; count++)
+            elements.push(symbol);
+    }
+    return elements;
+};
+// Writes the buffer to h5wasm's in-memory FS under a unique temp name, opens
+// it, runs the callback (awaiting async ones so cleanup can't race reads),
+// and always closes + unlinks afterwards.
+export async function with_h5_file(buffer, filename, callback) {
+    const { FS } = await h5wasm.ready;
+    const file_basename = filename
+        ?.split(`/`)
+        .at(-1)
+        ?.replaceAll(/[^\w.-]/g, `_`) ?? `temp`;
+    const unique_suffix = `${Date.now()}-${Math.random().toString(36).slice(2)}`;
+    const temp_filename = `${file_basename}-${unique_suffix}.h5`;
+    FS.writeFile(temp_filename, new Uint8Array(buffer));
+    let h5_file = null;
+    try {
+        h5_file = new h5wasm.File(temp_filename, `r`);
+        return await callback(h5_file);
+    }
+    finally {
+        h5_file?.close();
+        try {
+            FS.unlink(temp_filename);
+        }
+        catch {
+            /* temp file cleanup is best-effort */
+        }
+    }
+}

package/dist/trajectory/parse/hdf5.d.ts

@@ -1,2 +1,2 @@
 import type { TrajectoryType } from '../index';
-export declare function parse_torch_sim_hdf5(buffer: ArrayBuffer, filename?: string): Promise<TrajectoryType>;
+export declare function parse_hdf5_trajectory(buffer: ArrayBuffer, filename?: string): Promise<TrajectoryType>;