-
Notifications
You must be signed in to change notification settings - Fork 3.2k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Revert of More thorough overlapping cursor tests. (patchset #2 id:600…
…01 of https://codereview.chromium.org/2781623008/ ) Reason for revert: external/wpt/IndexedDB/parallel-overlapping-cursors.html is failing on the win7 webkit bot: https://build.chromium.org/p/chromium.webkit/builders/WebKit%20Win7%20%28dbg%29/builds/9419. Original issue's description: > More thorough overlapping cursor tests. > > BUG=705837 > > Review-Url: https://codereview.chromium.org/2781623008 > Cr-Commit-Position: refs/heads/master@{#461932} > Committed: https://chromium.googlesource.com/chromium/src/+/0c187ee3c2b8fc9d0dd81758f892cf6fd0ba4c3f [email protected],[email protected],[email protected] # Skipping CQ checks because original CL landed less than 1 days ago. NOPRESUBMIT=true NOTREECHECKS=true NOTRY=true BUG=705837 Review-Url: https://codereview.chromium.org/2803563002 Cr-Commit-Position: refs/heads/master@{#461991}
- Loading branch information
maxmorin
authored and
Unknown
committed
Apr 5, 2017
1 parent
a3767c3
commit ec42875
Showing
4 changed files
with
155 additions
and
330 deletions.
There are no files selected for viewing
This file was deleted.
Oops, something went wrong.
This file contains 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
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,37 +1,179 @@ | ||
<!doctype html> | ||
<meta charset="utf-8"> | ||
<meta name="timeout" content="long"> | ||
<title>IndexedDB: Interleaved iteration of multiple cursors over disjoint data ranges</title> | ||
<title>IndexedDB: Interleaved iteration of multiple cursors</title> | ||
<link rel="author" href="[email protected]" title="Victor Costan"> | ||
<script src="/resources/testharness.js"></script> | ||
<script src="/resources/testharnessreport.js"></script> | ||
<script src="support-promises.js"></script> | ||
<script src="interleaved-cursors-support.js"></script> | ||
<script> | ||
'use strict'; | ||
|
||
// Number of objects that each iterator goes over. | ||
const itemCount = 8; | ||
const itemCount = 10; | ||
|
||
// Ratio of small objects to large objects. | ||
const largeObjectRatio = 3; | ||
const largeObjectRatio = 5; | ||
|
||
// Size of large objects. This should exceed the size of a block in the storage | ||
// method underlying the browser's IndexedDB implementation. For example, this | ||
// needs to exceed the LevelDB block size on Chrome, and the SQLite block size | ||
// on Firefox. | ||
const largeObjectSize = 48 * 1024; | ||
|
||
function objectKey(cursorIndex, itemIndex) { | ||
const cursorString = cursorIndex.toString().padStart(5, '0'); | ||
const itemString = itemIndex.toString().padStart(5, '0'); | ||
return `${cursorString}-key-${itemString}`; | ||
return `${cursorIndex}-key-${itemIndex}`; | ||
} | ||
|
||
function objectValue(cursorIndex, itemIndex) { | ||
if ((cursorIndex * itemCount + itemIndex) % largeObjectRatio === 0) | ||
return largeObjectValue(cursorIndex, itemIndex); | ||
if ((cursorIndex * itemCount + itemIndex) % largeObjectRatio === 0) { | ||
// We use a typed array (as opposed to a string) because IndexedDB | ||
// implementations may serialize strings using UTF-8 or UTF-16, yielding | ||
// larger IndexedDB entries than we'd expect. It's very unlikely that an | ||
// IndexedDB implementation would use anything other than the raw buffer to | ||
// serialize a typed array. | ||
const buffer = new Uint8Array(largeObjectSize); | ||
|
||
// Some IndexedDB implementations, like LevelDB, compress their data blocks | ||
// before storing them to disk. We use a simple 32-bit xorshift PRNG, which | ||
// should be sufficient to foil any fast generic-purpose compression scheme. | ||
|
||
// 32-bit xorshift - the seed can't be zero | ||
let state = 1000 + (cursorIndex * itemCount + itemIndex); | ||
|
||
for (let i = 0; i < largeObjectSize; ++i) { | ||
state ^= state << 13; | ||
state ^= state >> 17; | ||
state ^= state << 5; | ||
buffer[i] = state & 0xff; | ||
} | ||
|
||
return buffer; | ||
} | ||
return [cursorIndex, 'small', itemIndex]; | ||
} | ||
|
||
// Writes the objects to be read by one cursor. Returns a promise that resolves | ||
// when the write completes. | ||
// | ||
// We want to avoid creating a large transaction, because that is outside the | ||
// test's scope, and it's a bad practice. So we break up the writes across | ||
// multiple transactions. For simplicity, each transaction writes all the | ||
// objects that will be read by a cursor. | ||
function writeCursorObjects(database, cursorIndex) { | ||
return new Promise((resolve, reject) => { | ||
const transaction = database.transaction('cache', 'readwrite'); | ||
transaction.onabort = () => { reject(transaction.error); }; | ||
|
||
const store = transaction.objectStore('cache'); | ||
for (let i = 0; i < itemCount; ++i) { | ||
store.put({ | ||
key: objectKey(cursorIndex, i), value: objectValue(cursorIndex, i)}); | ||
} | ||
transaction.oncomplete = resolve; | ||
}); | ||
} | ||
|
||
// Returns a promise that resolves when the store has been populated. | ||
function populateTestStore(testCase, database, cursorCount) { | ||
let promiseChain = Promise.resolve(); | ||
|
||
for (let i = 0; i < cursorCount; ++i) | ||
promiseChain = promiseChain.then(() => writeCursorObjects(database, i)); | ||
|
||
return promiseChain; | ||
} | ||
|
||
// Reads cursors in an interleaved fashion, as shown below. | ||
// | ||
// Given N cursors, each of which points to the beginning of a K-item sequence, | ||
// the following accesses will be made. | ||
// | ||
// OC(i) = open cursor i | ||
// RD(i, j) = read result of cursor i, which should be at item j | ||
// CC(i) = continue cursor i | ||
// | = wait for onsuccess on the previous OC or CC | ||
// | ||
// OC(1) | RD(1, 1) OC(2) | RD(2, 1) OC(3) | ... | RD(n-1, 1) CC(n) | | ||
// RD(n, 1) CC(1) | RD(1, 2) CC(2) | RD(2, 2) CC(3) | ... | RD(n-1, 2) CC(n) | | ||
// RD(n, 2) CC(1) | RD(1, 3) CC(2) | RD(2, 3) CC(3) | ... | RD(n-1, 3) CC(n) | | ||
// ... | ||
// RD(n, k-1) CC(1) | RD(1, k) CC(2) | RD(2, k) CC(3) | ... | RD(n-1, k) CC(n) | | ||
// RD(n, k) done | ||
function interleaveCursors(testCase, store, cursorCount) { | ||
return new Promise((resolve, reject) => { | ||
// The cursors used for iteration are stored here so each cursor's onsuccess | ||
// handler can call continue() on the next cursor. | ||
const cursors = []; | ||
|
||
// The results of IDBObjectStore.openCursor() calls are stored here so we | ||
// we can change the requests' onsuccess handler after every | ||
// IDBCursor.continue() call. | ||
const requests = []; | ||
|
||
const checkCursorState = (cursorIndex, itemIndex) => { | ||
const cursor = cursors[cursorIndex]; | ||
assert_equals(cursor.key, objectKey(cursorIndex, itemIndex)); | ||
assert_equals(cursor.value.key, objectKey(cursorIndex, itemIndex)); | ||
assert_equals( | ||
cursor.value.value.join('-'), | ||
objectValue(cursorIndex, itemIndex).join('-')); | ||
}; | ||
|
||
const openCursor = (cursorIndex, callback) => { | ||
const request = store.openCursor( | ||
IDBKeyRange.lowerBound(objectKey(cursorIndex, 0))); | ||
requests[cursorIndex] = request; | ||
|
||
request.onsuccess = testCase.step_func(() => { | ||
const cursor = request.result; | ||
cursors[cursorIndex] = cursor; | ||
checkCursorState(cursorIndex, 0); | ||
callback(); | ||
}); | ||
request.onerror = event => reject(request.error); | ||
}; | ||
|
||
const readItemFromCursor = (cursorIndex, itemIndex, callback) => { | ||
const request = requests[cursorIndex]; | ||
request.onsuccess = testCase.step_func(() => { | ||
const cursor = request.result; | ||
cursors[cursorIndex] = cursor; | ||
checkCursorState(cursorIndex, itemIndex); | ||
callback(); | ||
}); | ||
|
||
const cursor = cursors[cursorIndex]; | ||
cursor.continue(); | ||
}; | ||
|
||
// We open all the cursors one at a time, then cycle through the cursors and | ||
// call continue() on each of them. This access pattern causes maximal | ||
// trashing to an LRU cursor cache. Eviction scheme aside, any cache will | ||
// have to evict some cursors, and this access pattern verifies that the | ||
// cache correctly restores the state of evicted cursors. | ||
const steps = []; | ||
for (let cursorIndex = 0; cursorIndex < cursorCount; ++cursorIndex) | ||
steps.push(openCursor.bind(null, cursorIndex)); | ||
for (let itemIndex = 1; itemIndex < itemCount; ++itemIndex) { | ||
for (let cursorIndex = 0; cursorIndex < cursorCount; ++cursorIndex) | ||
steps.push(readItemFromCursor.bind(null, cursorIndex, itemIndex)); | ||
} | ||
|
||
const runStep = (stepIndex) => { | ||
if (stepIndex === steps.length) { | ||
resolve(); | ||
return; | ||
} | ||
steps[stepIndex](() => { runStep(stepIndex + 1); }); | ||
}; | ||
runStep(0); | ||
}); | ||
} | ||
|
||
for (let cursorCount of [1, 10, 100, 500]) { | ||
promise_test(testCase => { | ||
return createDatabase(testCase, (database, transaction) => { | ||
const store = database.createObjectStore('cache', { keyPath: 'key' }); | ||
const store = database.createObjectStore('cache', | ||
{ keyPath: 'key', autoIncrement: true }); | ||
}).then(database => { | ||
return populateTestStore(testCase, database, cursorCount).then( | ||
() => database); | ||
|
@@ -44,12 +186,11 @@ | |
transaction.onabort = () => { reject(transaction.error); }; | ||
|
||
const store = transaction.objectStore('cache'); | ||
return interleaveCursors(testCase, store, cursorCount, itemCount).then( | ||
return interleaveCursors(testCase, store, cursorCount).then( | ||
() => database); | ||
}).then(database => { | ||
database.close(); | ||
}); | ||
}, `${cursorCount} cursors`); | ||
} | ||
|
||
</script> |
Oops, something went wrong.