osobhandClaude Opus 5.5
42894bf93b
BTreeV2Header::parse_in, collect_btree_v2_records_in and find_btree_v2_records_in read one bounded window per node (its size is known from the parent before the node is read; a count stretched past node_size is checked against the end of the file first), with the whole-file bounds errors unchanged. With them, dense attributes, a SOHM B-tree index and huge fractal-heap objects no longer answer ContiguousStorageRequired, and group_v1/group_v2 listings, lookups and path resolution get *_in cores (resolve_group_children_in, resolve_child_in, resolve_path_any_in, ...). The &[u8] functions are thin wrappers, as in M1. The equivalence harness now fails on any ContiguousStorageRequired and compares v2 B-tree headers, records and descents, group listings, child lookups and paths; a unit test compares a two-level tree through a read_at-only storage truncated at every length and with every node byte flipped. Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
696 lines
28 KiB
Rust
696 lines
28 KiB
Rust
//! Equivalence harness for the range-read migration
|
|
//! (`docs/design/range-reads.md`, milestones M1 and M2).
|
|
//!
|
|
//! Every parser converted to [`Storage`] must give exactly what its `&[u8]`
|
|
//! form gives. This walks real files — the fixtures, files h5py writes to
|
|
//! exercise the less common structures, and optionally the conformance
|
|
//! corpus — and, for every object, runs each converted parser twice: over
|
|
//! the file as a slice, and over a [`CountingStorage`] that serves the same
|
|
//! bytes through `read_at` only (`as_contiguous()` is `None`, so no parser
|
|
//! can fall back to the whole slice). The results must be identical, value
|
|
//! for value and error for error.
|
|
//!
|
|
//! Nothing may answer [`FormatError::ContiguousStorageRequired`] any more:
|
|
//! since milestone M2 every read path works over `read_at` alone, the v2
|
|
//! B-tree structures (dense groups and attributes, a SOHM B-tree index,
|
|
//! huge fractal-heap objects) included.
|
|
//!
|
|
//! - `CLAWHDF5_STORAGE_CORPUS=dir[:dir...]` adds every `.h5`/`.hdf5`/`.he5`/
|
|
//! `.nc`/`.h5ad` file under those directories (the conformance corpus is
|
|
//! `conformance/.cache/corpus`); `CLAWHDF5_STORAGE_REPORT=1` prints the
|
|
//! per-file read counts.
|
|
//! - The h5py-written files honour `CLAWHDF5_PYTHON` and
|
|
//! `CLAWHDF5_REQUIRE_INTEROP` like the facade's interop tests.
|
|
|
|
use std::collections::{HashSet, VecDeque};
|
|
use std::fmt::Debug;
|
|
use std::path::{Path, PathBuf};
|
|
use std::process::Command;
|
|
|
|
use clawhdf5_format::attribute::{
|
|
extract_attributes_full, extract_attributes_full_in, extract_attributes_tolerant,
|
|
extract_attributes_tolerant_in,
|
|
};
|
|
use clawhdf5_format::attribute_info::AttributeInfoMessage;
|
|
use clawhdf5_format::btree_v1::{collect_symbol_table_nodes, collect_symbol_table_nodes_in};
|
|
use clawhdf5_format::btree_v2::{
|
|
BTreeV2Header, collect_btree_v2_records, collect_btree_v2_records_in, find_btree_v2_records,
|
|
find_btree_v2_records_in,
|
|
};
|
|
use clawhdf5_format::data_layout::DataLayout;
|
|
use clawhdf5_format::dataspace::Dataspace;
|
|
use clawhdf5_format::datatype::Datatype;
|
|
use clawhdf5_format::error::FormatError;
|
|
use clawhdf5_format::extensible_array::{
|
|
ExtensibleArrayHeader, read_extensible_array_chunks, read_extensible_array_chunks_in,
|
|
};
|
|
use clawhdf5_format::fill_value::{dataset_fill_value_from_storage, dataset_fill_value_in};
|
|
use clawhdf5_format::fixed_array::{
|
|
FixedArrayHeader, read_fixed_array_chunks, read_fixed_array_chunks_in,
|
|
};
|
|
use clawhdf5_format::fractal_heap::FractalHeapHeader;
|
|
use clawhdf5_format::group_v2;
|
|
use clawhdf5_format::link_info::LinkInfoMessage;
|
|
use clawhdf5_format::local_heap::LocalHeap;
|
|
use clawhdf5_format::message_type::MessageType;
|
|
use clawhdf5_format::object_header::ObjectHeader;
|
|
use clawhdf5_format::shared_message::{
|
|
self, load_sohm_table, load_sohm_table_in, message_data_with_sohm, message_data_with_sohm_in,
|
|
parse_sohm_btree_entries, parse_sohm_btree_entries_in, parse_sohm_list, parse_sohm_list_in,
|
|
};
|
|
use clawhdf5_format::signature::split_user_block;
|
|
use clawhdf5_format::storage::{CountingStorage, Storage};
|
|
use clawhdf5_format::superblock::Superblock;
|
|
use clawhdf5_format::superblock_ext::{
|
|
cache_image_state, cache_image_state_in, read_superblock_extension,
|
|
read_superblock_extension_in,
|
|
};
|
|
use clawhdf5_format::symbol_table::{SymbolTableMessage, SymbolTableNode};
|
|
|
|
/// Objects visited per file, heap objects read per heap: enough to cover
|
|
/// every structure kind while keeping a 35 000-group file fast.
|
|
const MAX_OBJECTS: usize = 1500;
|
|
const MAX_HEAP_IDS: usize = 200;
|
|
|
|
#[derive(Default, Debug)]
|
|
struct Tally {
|
|
files: usize,
|
|
objects: usize,
|
|
checks: usize,
|
|
reads: u64,
|
|
bytes: u64,
|
|
/// Chunk indexes (fixed and extensible arrays) read, and the most bytes
|
|
/// one of them took through the storage.
|
|
chunk_indexes: usize,
|
|
max_chunk_index_bytes: u64,
|
|
}
|
|
|
|
struct Walk<'a> {
|
|
slice: &'a [u8],
|
|
storage: &'a CountingStorage,
|
|
name: String,
|
|
tally: &'a mut Tally,
|
|
}
|
|
|
|
impl Walk<'_> {
|
|
/// The storage result must equal the slice result; no parser may ask
|
|
/// for the whole file.
|
|
fn same<T: Debug>(
|
|
&mut self,
|
|
what: &str,
|
|
want: &Result<T, FormatError>,
|
|
got: &Result<T, FormatError>,
|
|
) {
|
|
self.tally.checks += 1;
|
|
if let Err(FormatError::ContiguousStorageRequired(site)) = got {
|
|
panic!(
|
|
"{}: {what} fell back to the whole file ({site}); every read path \
|
|
must work through read_at",
|
|
self.name
|
|
);
|
|
}
|
|
let (w, g) = (format!("{want:?}"), format!("{got:?}"));
|
|
assert!(
|
|
w == g,
|
|
"{}: {what} differs\n slice: {}\n storage: {}",
|
|
self.name,
|
|
&w[..w.len().min(600)],
|
|
&g[..g.len().min(600)]
|
|
);
|
|
}
|
|
|
|
fn index_read(&mut self, bytes_before: u64) {
|
|
self.tally.chunk_indexes += 1;
|
|
let bytes = self.storage.bytes_read() - bytes_before;
|
|
self.tally.max_chunk_index_bytes = self.tally.max_chunk_index_bytes.max(bytes);
|
|
}
|
|
|
|
fn st(&self) -> &dyn Storage {
|
|
self.storage
|
|
}
|
|
|
|
fn run(&mut self) {
|
|
let slice = self.slice;
|
|
let sb = Superblock::parse(slice, 0);
|
|
self.same("superblock", &sb, &Superblock::parse_in(self.st(), 0));
|
|
let Ok(sb) = sb else { return };
|
|
let (os, ls) = (sb.offset_size, sb.length_size);
|
|
|
|
let want = read_superblock_extension(slice, &sb);
|
|
self.same(
|
|
"superblock extension",
|
|
&want,
|
|
&read_superblock_extension_in(self.st(), &sb),
|
|
);
|
|
let want = cache_image_state(slice, &sb);
|
|
self.same("cache image", &want, &cache_image_state_in(self.st(), &sb));
|
|
let table = load_sohm_table(slice, os, ls);
|
|
self.same("SOHM table", &table, &load_sohm_table_in(self.st(), os, ls));
|
|
if let Ok(Some(table)) = &table {
|
|
for idx in &table.indexes {
|
|
if idx.index_type == 0 {
|
|
let want =
|
|
parse_sohm_list(slice, idx.index_addr as usize, idx.num_messages, os);
|
|
let got = parse_sohm_list_in(self.st(), idx.index_addr, idx.num_messages, os);
|
|
self.same("SOHM list", &want, &got);
|
|
} else {
|
|
let want = parse_sohm_btree_entries(slice, idx.index_addr as usize, os, ls);
|
|
let got = parse_sohm_btree_entries_in(self.st(), idx.index_addr, os, ls);
|
|
self.same("SOHM B-tree", &want, &got);
|
|
}
|
|
}
|
|
}
|
|
|
|
let mut seen = HashSet::new();
|
|
let mut queue = VecDeque::from([sb.root_group_address]);
|
|
while let Some(addr) = queue.pop_front() {
|
|
if seen.len() >= MAX_OBJECTS || !seen.insert(addr) {
|
|
continue;
|
|
}
|
|
self.tally.objects += 1;
|
|
self.check_object(&sb, addr);
|
|
let children = group_v2::resolve_group_children(slice, &sb, addr);
|
|
let got = group_v2::resolve_group_children_in(self.st(), &sb, addr);
|
|
self.same("group listing", &children, &got);
|
|
if let Ok(children) = children {
|
|
for c in children.iter().take(MAX_HEAP_IDS) {
|
|
let want = group_v2::resolve_child(slice, &sb, addr, &c.name);
|
|
let got = group_v2::resolve_child_in(self.st(), &sb, addr, &c.name);
|
|
self.same("child lookup", &want, &got);
|
|
}
|
|
// A name no group has: the lookup's not-found path.
|
|
let want = group_v2::resolve_child(slice, &sb, addr, "no such child");
|
|
let got = group_v2::resolve_child_in(self.st(), &sb, addr, "no such child");
|
|
self.same("child lookup (missing)", &want, &got);
|
|
queue.extend(children.iter().map(|c| c.object_header_address));
|
|
}
|
|
}
|
|
// Paths: every listed name from the root, and one that is missing.
|
|
if let Ok(children) = group_v2::resolve_group_children(slice, &sb, sb.root_group_address) {
|
|
for c in children.iter().take(MAX_HEAP_IDS) {
|
|
let path = format!("/{}", c.name);
|
|
let want = group_v2::resolve_path_any(slice, &sb, &path);
|
|
let got = group_v2::resolve_path_any_in(self.st(), &sb, &path);
|
|
self.same("path", &want, &got);
|
|
}
|
|
}
|
|
let want = group_v2::resolve_path_any(slice, &sb, "/no/such/path");
|
|
let got = group_v2::resolve_path_any_in(self.st(), &sb, "/no/such/path");
|
|
self.same("path (missing)", &want, &got);
|
|
}
|
|
|
|
fn check_object(&mut self, sb: &Superblock, addr: u64) {
|
|
let slice = self.slice;
|
|
let (os, ls) = (sb.offset_size, sb.length_size);
|
|
let header = ObjectHeader::parse(slice, addr as usize, os, ls);
|
|
self.same(
|
|
"object header",
|
|
&header,
|
|
&ObjectHeader::parse_in(self.st(), addr, os, ls),
|
|
);
|
|
let Ok(header) = header else { return };
|
|
|
|
let want = extract_attributes_full(slice, &header, os, ls);
|
|
self.same(
|
|
"attributes",
|
|
&want,
|
|
&extract_attributes_full_in(self.st(), &header, os, ls),
|
|
);
|
|
let want = extract_attributes_tolerant(slice, &header, os, ls);
|
|
let got = extract_attributes_tolerant_in(self.st(), &header, os, ls);
|
|
self.same("attributes (tolerant)", &want, &got);
|
|
let want = dataset_fill_value_in(slice, &header.messages, os, ls);
|
|
self.same(
|
|
"fill value",
|
|
&want,
|
|
&dataset_fill_value_from_storage(self.st(), &header.messages, os, ls),
|
|
);
|
|
|
|
for msg in &header.messages {
|
|
if shared_message::is_shared(msg.flags) {
|
|
let want = message_data_with_sohm(slice, msg, os, ls);
|
|
let got = message_data_with_sohm_in(self.st(), msg, os, ls);
|
|
self.same("shared message", &want, &got);
|
|
}
|
|
match msg.msg_type {
|
|
MessageType::SymbolTable => {
|
|
if let Ok(stm) = SymbolTableMessage::parse(&msg.data, os) {
|
|
self.check_v1_group(&stm, os, ls);
|
|
}
|
|
}
|
|
MessageType::LinkInfo => {
|
|
if let Ok(li) = LinkInfoMessage::parse(&msg.data, os) {
|
|
self.check_heap(
|
|
li.fractal_heap_address,
|
|
li.btree_name_index_address,
|
|
4,
|
|
os,
|
|
ls,
|
|
);
|
|
}
|
|
}
|
|
MessageType::AttributeInfo => {
|
|
if let Ok(ai) = AttributeInfoMessage::parse(&msg.data, os) {
|
|
self.check_heap(
|
|
ai.fractal_heap_address,
|
|
ai.btree_name_index_address,
|
|
0,
|
|
os,
|
|
ls,
|
|
);
|
|
}
|
|
}
|
|
_ => {}
|
|
}
|
|
}
|
|
self.check_layout(&header, os, ls);
|
|
}
|
|
|
|
/// A symbol-table group: its local heap, B-tree, nodes and names.
|
|
fn check_v1_group(&mut self, stm: &SymbolTableMessage, os: u8, ls: u8) {
|
|
let slice = self.slice;
|
|
let heap = LocalHeap::parse(slice, stm.local_heap_address as usize, os, ls);
|
|
self.same(
|
|
"local heap",
|
|
&heap,
|
|
&LocalHeap::parse_in(self.st(), stm.local_heap_address, os, ls),
|
|
);
|
|
let nodes = collect_symbol_table_nodes(slice, stm.btree_address, os, ls);
|
|
let got = collect_symbol_table_nodes_in(self.st(), stm.btree_address, os, ls);
|
|
self.same("group B-tree", &nodes, &got);
|
|
let Ok(heap) = heap else { return };
|
|
let want = heap.validate_free_list(slice, ls);
|
|
self.same(
|
|
"local heap free list",
|
|
&want,
|
|
&heap.validate_free_list_in(self.st(), ls),
|
|
);
|
|
let Ok(nodes) = nodes else { return };
|
|
for &node in nodes.iter().take(MAX_HEAP_IDS) {
|
|
let snod = SymbolTableNode::parse(slice, node as usize, os);
|
|
self.same(
|
|
"symbol table node",
|
|
&snod,
|
|
&SymbolTableNode::parse_in(self.st(), node, os),
|
|
);
|
|
let Ok(snod) = snod else { continue };
|
|
for e in &snod.entries {
|
|
let want = heap.read_string(slice, e.link_name_offset);
|
|
self.same(
|
|
"link name",
|
|
&want,
|
|
&heap.read_string_in(self.st(), e.link_name_offset),
|
|
);
|
|
}
|
|
}
|
|
}
|
|
|
|
/// A dense group's or dense attributes' fractal heap: the header, and
|
|
/// the objects its name index points at. `id_at` is where the heap ID
|
|
/// starts in a name-index record (after the hash for links).
|
|
fn check_heap(&mut self, heap: Option<u64>, index: Option<u64>, id_at: usize, os: u8, ls: u8) {
|
|
let slice = self.slice;
|
|
let Some(heap_addr) = heap else { return };
|
|
let fh = FractalHeapHeader::parse(slice, heap_addr as usize, os, ls);
|
|
self.same(
|
|
"fractal heap",
|
|
&fh,
|
|
&FractalHeapHeader::parse_in(self.st(), heap_addr, os, ls),
|
|
);
|
|
let (Ok(fh), Some(index)) = (fh, index) else {
|
|
return;
|
|
};
|
|
let bt = BTreeV2Header::parse(slice, index as usize, os, ls);
|
|
self.same(
|
|
"v2 B-tree header",
|
|
&bt,
|
|
&BTreeV2Header::parse_in(self.st(), index, os, ls),
|
|
);
|
|
let Ok(bt) = bt else { return };
|
|
let records = collect_btree_v2_records(slice, &bt, os, ls);
|
|
let got = collect_btree_v2_records_in(self.st(), &bt, os, ls);
|
|
self.same("v2 B-tree records", &records, &got);
|
|
let Ok(records) = records else { return };
|
|
// Descents to single records (by their bytes), as name lookups do.
|
|
for rec in records.iter().take(8) {
|
|
let key = rec.data.clone();
|
|
let want = find_btree_v2_records(slice, &bt, os, &mut |r| r.cmp(&key[..]));
|
|
let got = find_btree_v2_records_in(self.st(), &bt, os, &mut |r| r.cmp(&key[..]));
|
|
self.same("v2 B-tree descent", &want, &got);
|
|
}
|
|
let id_len = fh.heap_id_length as usize;
|
|
for rec in records.iter().take(MAX_HEAP_IDS) {
|
|
let Some(id) = rec.data.get(id_at..id_at + id_len) else {
|
|
continue;
|
|
};
|
|
let want = fh.read_managed_object(slice, id, os);
|
|
self.same(
|
|
"heap object",
|
|
&want,
|
|
&fh.read_managed_object_in(self.st(), id, os),
|
|
);
|
|
}
|
|
}
|
|
|
|
/// A dataset's layout: VDS mappings, and fixed/extensible array chunk
|
|
/// indexes.
|
|
fn check_layout(&mut self, header: &ObjectHeader, os: u8, ls: u8) {
|
|
let slice = self.slice;
|
|
let find = |t: MessageType| {
|
|
header
|
|
.messages
|
|
.iter()
|
|
.find(|m| m.msg_type == t)
|
|
.and_then(|m| shared_message::message_data_with_sohm(slice, m, os, ls).ok())
|
|
};
|
|
let Some(layout) = find(MessageType::DataLayout) else {
|
|
return;
|
|
};
|
|
let Ok(layout) = DataLayout::parse(&layout, os, ls) else {
|
|
return;
|
|
};
|
|
match &layout {
|
|
DataLayout::Virtual { .. } => {
|
|
let (mut want, mut got) = (layout.clone(), layout.clone());
|
|
let w = want.resolve_vds_mappings(slice, ls).map(|()| want);
|
|
let g = got.resolve_vds_mappings_in(self.st(), ls).map(|()| got);
|
|
self.same("VDS mappings", &w, &g);
|
|
}
|
|
DataLayout::Chunked {
|
|
chunk_dimensions,
|
|
btree_address: Some(addr),
|
|
version: 4,
|
|
chunk_index_type: Some(kind @ (3 | 4)),
|
|
..
|
|
} => {
|
|
let (Some(ds), Some(dt)) =
|
|
(find(MessageType::Dataspace), find(MessageType::Datatype))
|
|
else {
|
|
return;
|
|
};
|
|
let (Ok(ds), Ok((dt, _))) = (Dataspace::parse(&ds, ls), Datatype::parse(&dt))
|
|
else {
|
|
return;
|
|
};
|
|
let rank = ds.dimensions.len();
|
|
if chunk_dimensions.len() < rank {
|
|
return;
|
|
}
|
|
let dims = &chunk_dimensions[..rank];
|
|
let max = ds.max_dimensions.as_deref();
|
|
let es = dt.type_size();
|
|
if *kind == 3 {
|
|
let h = FixedArrayHeader::parse(slice, *addr as usize, os, ls);
|
|
self.same(
|
|
"fixed array header",
|
|
&h,
|
|
&FixedArrayHeader::parse_in(self.st(), *addr, os, ls),
|
|
);
|
|
let Ok(h) = h else { return };
|
|
let want =
|
|
read_fixed_array_chunks(slice, &h, &ds.dimensions, max, dims, es, os, ls);
|
|
let before = self.storage.bytes_read();
|
|
let got = read_fixed_array_chunks_in(
|
|
self.st(),
|
|
&h,
|
|
&ds.dimensions,
|
|
max,
|
|
dims,
|
|
es,
|
|
os,
|
|
ls,
|
|
);
|
|
self.index_read(before);
|
|
self.same("fixed array chunks", &want, &got);
|
|
} else {
|
|
let h = ExtensibleArrayHeader::parse(slice, *addr as usize, os, ls);
|
|
let got = ExtensibleArrayHeader::parse_in(self.st(), *addr, os, ls);
|
|
self.same("extensible array header", &h, &got);
|
|
let Ok(h) = h else { return };
|
|
let want = read_extensible_array_chunks(
|
|
slice,
|
|
&h,
|
|
&ds.dimensions,
|
|
max,
|
|
dims,
|
|
es,
|
|
os,
|
|
ls,
|
|
);
|
|
let before = self.storage.bytes_read();
|
|
let got = read_extensible_array_chunks_in(
|
|
self.st(),
|
|
&h,
|
|
&ds.dimensions,
|
|
max,
|
|
dims,
|
|
es,
|
|
os,
|
|
ls,
|
|
);
|
|
self.index_read(before);
|
|
self.same("extensible array chunks", &want, &got);
|
|
}
|
|
}
|
|
_ => {}
|
|
}
|
|
}
|
|
}
|
|
|
|
fn check_file(path: &Path, tally: &mut Tally) {
|
|
let Ok(bytes) = std::fs::read(path) else {
|
|
return;
|
|
};
|
|
check_bytes(&path.display().to_string(), &bytes, tally);
|
|
}
|
|
|
|
fn check_bytes(name: &str, bytes: &[u8], tally: &mut Tally) {
|
|
let Ok((_, hdf5)) = split_user_block(bytes) else {
|
|
return;
|
|
};
|
|
let storage = CountingStorage::new(hdf5.to_vec());
|
|
let before = (tally.checks, tally.objects);
|
|
let mut walk = Walk {
|
|
slice: hdf5,
|
|
storage: &storage,
|
|
name: name.to_string(),
|
|
tally,
|
|
};
|
|
walk.run();
|
|
tally.files += 1;
|
|
tally.reads += storage.reads();
|
|
tally.bytes += storage.bytes_read();
|
|
if std::env::var("CLAWHDF5_STORAGE_REPORT").is_ok_and(|v| v == "1") {
|
|
eprintln!(
|
|
"{:>6} objects {:>7} checks {:>8} reads {:>12} bytes {}",
|
|
tally.objects - before.1,
|
|
tally.checks - before.0,
|
|
storage.reads(),
|
|
storage.bytes_read(),
|
|
name
|
|
);
|
|
}
|
|
}
|
|
|
|
fn hdf5_files(dir: &Path, out: &mut Vec<PathBuf>) {
|
|
let Ok(entries) = std::fs::read_dir(dir) else {
|
|
return;
|
|
};
|
|
for e in entries.flatten() {
|
|
let p = e.path();
|
|
if p.is_dir() {
|
|
hdf5_files(&p, out);
|
|
} else if p
|
|
.extension()
|
|
.and_then(|x| x.to_str())
|
|
.is_some_and(|x| matches!(x, "h5" | "hdf5" | "he5" | "nc" | "h5ad" | "hdf"))
|
|
{
|
|
out.push(p);
|
|
}
|
|
}
|
|
}
|
|
|
|
#[test]
|
|
fn fixtures_parse_identically_through_storage() {
|
|
let mut files = Vec::new();
|
|
hdf5_files(
|
|
&Path::new(env!("CARGO_MANIFEST_DIR")).join("tests/fixtures"),
|
|
&mut files,
|
|
);
|
|
files.sort();
|
|
assert!(files.len() >= 40, "{} fixtures", files.len());
|
|
let mut tally = Tally::default();
|
|
for f in &files {
|
|
check_file(f, &mut tally);
|
|
}
|
|
eprintln!("fixtures: {tally:?}");
|
|
assert!(tally.objects >= 150 && tally.checks >= 1000, "{tally:?}");
|
|
// Reads really went through read_at.
|
|
assert!(tally.reads > tally.objects as u64);
|
|
}
|
|
|
|
#[test]
|
|
fn corpus_parses_identically_through_storage() {
|
|
let Ok(dirs) = std::env::var("CLAWHDF5_STORAGE_CORPUS") else {
|
|
eprintln!("CLAWHDF5_STORAGE_CORPUS not set; skipping the corpus");
|
|
return;
|
|
};
|
|
let mut files = Vec::new();
|
|
for d in std::env::split_paths(&dirs) {
|
|
hdf5_files(&d, &mut files);
|
|
}
|
|
files.sort();
|
|
let mut tally = Tally::default();
|
|
for f in &files {
|
|
check_file(f, &mut tally);
|
|
}
|
|
eprintln!("corpus: {tally:?}");
|
|
assert!(tally.files > 0);
|
|
}
|
|
|
|
fn python() -> String {
|
|
std::env::var("CLAWHDF5_PYTHON").unwrap_or_else(|_| "python3".to_string())
|
|
}
|
|
|
|
fn interop_required() -> bool {
|
|
std::env::var("CLAWHDF5_REQUIRE_INTEROP").is_ok_and(|v| v == "1")
|
|
}
|
|
|
|
/// Files h5py writes to cover what the fixtures do not: extensible arrays
|
|
/// deep enough for super blocks and paged data blocks, paged fixed arrays,
|
|
/// big symbol-table and dense groups, dense and shared attributes, a SOHM
|
|
/// list and a SOHM B-tree, committed datatypes, and a user block.
|
|
const GENERATE: &str = r#"
|
|
import ctypes, glob, os, sys, h5py, numpy as np
|
|
out = sys.argv[1]
|
|
def p(n): return os.path.join(out, n)
|
|
|
|
with h5py.File(p('ea.h5'), 'w', libver='latest') as f:
|
|
# One unlimited dimension: extensible array. 3000 chunks reach super
|
|
# blocks; 2-element chunks of int8 keep data small.
|
|
d = f.create_dataset('ea', shape=(6000,), maxshape=(None,), chunks=(2,), dtype='i1')
|
|
d[:] = np.arange(6000) % 100
|
|
d2 = f.create_dataset('ea_deflate', shape=(4000, 3), maxshape=(None, 3), chunks=(2, 3),
|
|
dtype='f4', compression='gzip')
|
|
d2[:] = np.random.default_rng(1).random((4000, 3))
|
|
d3 = f.create_dataset('ea_sparse', shape=(100000,), maxshape=(None,), chunks=(4,), dtype='i2')
|
|
d3[0:8] = 1; d3[50000:50004] = 2; d3[99996:] = 3
|
|
fa = f.create_dataset('fa_paged', shape=(5000,), chunks=(1,), dtype='u1')
|
|
fa[::3] = 7
|
|
fa2 = f.create_dataset('fa', shape=(40, 40), chunks=(8, 8), dtype='f8', compression='gzip')
|
|
fa2[:] = 1.5
|
|
for i in range(30):
|
|
f.attrs[f'a{i}'] = np.arange(i + 1)
|
|
g = f.create_group('dense')
|
|
for i in range(300):
|
|
g.create_group(f'child{i:04d}').attrs['i'] = i
|
|
f['committed'] = np.dtype([('x', 'i4'), ('y', 'f8')])
|
|
f.create_dataset('uses_committed', shape=(3,), dtype=f['committed'])
|
|
f['uses_committed'].attrs.create('ta', data=np.zeros(2, dtype=f['committed'].dtype), dtype=f['committed'])
|
|
f.attrs['vl'] = ['alpha', 'beta', 'gamma']
|
|
|
|
with h5py.File(p('v1_groups.h5'), 'w', libver='earliest', userblock_size=512) as f:
|
|
for i in range(400):
|
|
g = f.create_group(f'g{i:04d}')
|
|
g.attrs['n'] = i
|
|
f.create_dataset('x', data=np.arange(10))
|
|
|
|
# Paged chunk indexes whose data blocks are bigger than the storage reads
|
|
# in one piece (1 MiB), with two chunks written: a fixed array of 300 000
|
|
# chunks (a 2.4 MB data block) and an extensible array grown to 1.2e9 (its
|
|
# last data block holds 131 072 chunks: over 1 MiB).
|
|
with h5py.File(p('big_paged.h5'), 'w', libver='latest') as f:
|
|
d = f.create_dataset('fa', shape=(300000,), chunks=(1,), dtype='u1')
|
|
d[5] = 1; d[250000] = 2
|
|
e = f.create_dataset('ea', shape=(1,), maxshape=(None,), chunks=(1,), dtype='u1')
|
|
e.resize((1200000000,)); e[10] = 1; e[1100000000] = 3
|
|
|
|
libs = glob.glob(os.path.join(os.path.dirname(h5py.__file__), '..', 'h5py.libs', 'libhdf5-*.so*'))
|
|
if libs:
|
|
lib = ctypes.CDLL(libs[0])
|
|
lib.H5Pset_shared_mesg_nindexes.argtypes = [ctypes.c_int64, ctypes.c_uint]
|
|
lib.H5Pset_shared_mesg_index.argtypes = [ctypes.c_int64, ctypes.c_uint, ctypes.c_uint, ctypes.c_uint]
|
|
lib.H5Pset_shared_mesg_phase_change.argtypes = [ctypes.c_int64, ctypes.c_uint, ctypes.c_uint]
|
|
for name, list_max in [('sohm_list.h5', 50), ('sohm_btree.h5', 0)]:
|
|
fcpl = h5py.h5p.create(h5py.h5p.FILE_CREATE)
|
|
assert lib.H5Pset_shared_mesg_nindexes(fcpl.id, 1) >= 0
|
|
# datatype, dataspace, fill value, filter pipeline, attribute
|
|
assert lib.H5Pset_shared_mesg_index(fcpl.id, 0, 0x02 | 0x04 | 0x08 | 0x10 | 0x20, 1) >= 0
|
|
assert lib.H5Pset_shared_mesg_phase_change(fcpl.id, list_max, 0) >= 0
|
|
fapl = h5py.h5p.create(h5py.h5p.FILE_ACCESS)
|
|
fapl.set_libver_bounds(h5py.h5f.LIBVER_LATEST, h5py.h5f.LIBVER_LATEST)
|
|
fid = h5py.h5f.create(p(name).encode(), h5py.h5f.ACC_TRUNC, fcpl=fcpl, fapl=fapl)
|
|
with h5py.File(fid) as f:
|
|
for i in range(20):
|
|
d = f.create_dataset(f'd{i}', shape=(10, i + 1), dtype='f4', fillvalue=-1.0,
|
|
chunks=(5, 1), compression='gzip')
|
|
d.attrs['units'] = 'metres per second, a long enough string to share'
|
|
d.attrs['scale'] = np.arange(20, dtype='f8')
|
|
print('ok')
|
|
"#;
|
|
|
|
#[test]
|
|
fn h5py_files_parse_identically_through_storage() {
|
|
let dir = Path::new(env!("CARGO_TARGET_TMPDIR")).join("storage_equivalence");
|
|
let _ = std::fs::remove_dir_all(&dir);
|
|
std::fs::create_dir_all(&dir).unwrap();
|
|
let out = Command::new(python())
|
|
.args(["-c", GENERATE, dir.to_str().unwrap()])
|
|
.output();
|
|
match out {
|
|
Ok(o) if o.status.success() => {}
|
|
Ok(o) if interop_required() => panic!(
|
|
"h5py generation failed:\n{}\n{}",
|
|
String::from_utf8_lossy(&o.stdout),
|
|
String::from_utf8_lossy(&o.stderr)
|
|
),
|
|
Err(e) if interop_required() => panic!("python not available: {e}"),
|
|
_ => {
|
|
eprintln!("python3 with h5py unavailable; skipping");
|
|
return;
|
|
}
|
|
}
|
|
let mut files = Vec::new();
|
|
hdf5_files(&dir, &mut files);
|
|
files.sort();
|
|
let names: Vec<_> = files
|
|
.iter()
|
|
.map(|f| f.file_name().unwrap().to_string_lossy().into_owned())
|
|
.collect();
|
|
for want in ["ea.h5", "v1_groups.h5", "big_paged.h5"] {
|
|
assert!(names.iter().any(|n| n == want), "{names:?}");
|
|
}
|
|
if interop_required() {
|
|
assert!(names.iter().any(|n| n == "sohm_btree.h5"), "{names:?}");
|
|
}
|
|
let mut tally = Tally::default();
|
|
for f in &files {
|
|
if f.ends_with("big_paged.h5") {
|
|
// Only the pages in use are read, not the whole data blocks:
|
|
// read in one piece, the fixed array took 2.4 MB and the
|
|
// extensible array 1.2 MB (its super block's page bitmap and
|
|
// block addresses are most of what remains).
|
|
let mut big = Tally::default();
|
|
check_file(f, &mut big);
|
|
eprintln!("big paged blocks: {big:?}");
|
|
assert_eq!(big.chunk_indexes, 2, "{big:?}");
|
|
assert!(big.max_chunk_index_bytes < 256 << 10, "{big:?}");
|
|
// Truncated anywhere, the page-by-page reads still agree with
|
|
// the slice reads (errors included).
|
|
let bytes = std::fs::read(f).unwrap();
|
|
for cut in (0..bytes.len()).step_by(bytes.len() / 97) {
|
|
check_bytes(
|
|
&format!("big_paged.h5 cut at {cut}"),
|
|
&bytes[..cut],
|
|
&mut big,
|
|
);
|
|
}
|
|
eprintln!("big paged blocks, truncated: {big:?}");
|
|
assert!(big.chunk_indexes > 100, "{big:?}");
|
|
}
|
|
check_file(f, &mut tally);
|
|
}
|
|
eprintln!("h5py files: {tally:?}");
|
|
assert!(tally.objects >= 700, "{tally:?}");
|
|
}
|