@node-projects/speedscope 2.0.0 → 2.1.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +40 -2
- package/bin/cli.mjs +0 -0
- package/dist/release/index.html +2 -1
- package/dist/release/release.txt +3 -3
- package/dist/release/speedscope-OUBBGH23.js +191 -0
- package/dist/release/streamparser-json.LICENSE +21 -0
- package/docs/JAVA_THREAD_DUMPS.md +32 -0
- package/docs/LARGE_TRACE_PERFORMANCE.md +258 -6
- package/docs/OFFLINE.md +58 -0
- package/docs/PROFILE_FILTERS.md +73 -0
- package/docs/RELEASING.md +38 -9
- package/docs/SEARCH_AND_LAYOUT.md +260 -0
- package/docs/SOURCE_AND_LABELS.md +69 -0
- package/package.json +10 -2
- package/dist/release/favicon-16x16-V2DMIAZS.js.map +0 -7
- package/dist/release/favicon-32x32-THY3JDJL.js.map +0 -7
- package/dist/release/favicon-M34RF7BI.js.map +0 -7
- package/dist/release/speedscope-GHPHNKXC.css.map +0 -7
- package/dist/release/speedscope-IT6IHU5E.js +0 -188
- package/dist/release/speedscope-IT6IHU5E.js.map +0 -7
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
MIT License
|
|
2
|
+
|
|
3
|
+
Copyright (c) 2020 Juanjo Diaz
|
|
4
|
+
|
|
5
|
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
6
|
+
of this software and associated documentation files (the "Software"), to deal
|
|
7
|
+
in the Software without restriction, including without limitation the rights
|
|
8
|
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
9
|
+
copies of the Software, and to permit persons to whom the Software is
|
|
10
|
+
furnished to do so, subject to the following conditions:
|
|
11
|
+
|
|
12
|
+
The above copyright notice and this permission notice shall be included in all
|
|
13
|
+
copies or substantial portions of the Software.
|
|
14
|
+
|
|
15
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
16
|
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
17
|
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
18
|
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
19
|
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
20
|
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
21
|
+
SOFTWARE.
|
|
@@ -0,0 +1,32 @@
|
|
|
1
|
+
# Java thread-dump snapshots
|
|
2
|
+
|
|
3
|
+
Import traditional HotSpot text output from `jstack PID` or
|
|
4
|
+
`jcmd PID Thread.print` by opening a `.jstack` file or a text file containing
|
|
5
|
+
the `Full thread dump …` header. Concatenate captures from the same JVM to
|
|
6
|
+
compare observed stacks across repeated snapshots. Extract ZIP archives first.
|
|
7
|
+
|
|
8
|
+
Each observed Java stack contributes one raw sample. The importer creates an
|
|
9
|
+
aggregate profile per thread name and Java thread ID, falling back to `tid`
|
|
10
|
+
when the Java ID is absent. Renamed threads appear separately. Missing threads
|
|
11
|
+
and threads without Java frames add no samples. The profile selector switches
|
|
12
|
+
between threads; Left Heavy, Sandwich and Call Tree show the observations.
|
|
13
|
+
Time Order is unavailable because snapshots do not measure execution duration.
|
|
14
|
+
CPU, elapsed, timestamps, state and lock text are not converted into weights.
|
|
15
|
+
|
|
16
|
+
Frames retain method/module names and available source files and positive line
|
|
17
|
+
numbers. Native Method and Unknown Source entries keep their identity without
|
|
18
|
+
inventing a source location. Deadlock summaries are ignored because they repeat
|
|
19
|
+
stacks already listed in the snapshot. Malformed thread headers, malformed Java
|
|
20
|
+
frames and duplicate thread identities within a snapshot produce an import error.
|
|
21
|
+
|
|
22
|
+
The parser iterates lines and retains the current stack and current snapshot's
|
|
23
|
+
thread identities, rather than collecting every line or observation. The text
|
|
24
|
+
input and resulting profile model still occupy memory; this format does not use
|
|
25
|
+
the duration-trace streaming importer. Prefer captures from one JVM per file.
|
|
26
|
+
ZIP input, structured virtual-thread JSON, native address stacks and lock/state
|
|
27
|
+
visualization are outside the supported format.
|
|
28
|
+
|
|
29
|
+
The checked-in [synthetic example](../sample/profiles/java-thread-dump/synthetic.jstack)
|
|
30
|
+
contains repeated snapshots and source locations. Format references are
|
|
31
|
+
[OpenJDK's traditional dump examples](https://github.com/openjdk/jdk/blob/master/test/hotspot/jtreg/serviceability/tmtools/jstack/utils/DefaultFormat.java)
|
|
32
|
+
and the [Oracle Java troubleshooting guide](https://docs.oracle.com/en/java/javase/21/troubleshoot/troubleshooting-guide.pdf).
|
|
@@ -86,9 +86,261 @@ completed download, disabled Time Order for aggregates and combined-stack
|
|
|
86
86
|
inversion. The downloaded JSON was compared exactly with the standard exporter,
|
|
87
87
|
then reimported with all profile totals preserved.
|
|
88
88
|
|
|
89
|
-
|
|
90
|
-
|
|
91
|
-
|
|
92
|
-
|
|
93
|
-
|
|
94
|
-
and
|
|
89
|
+
## Worker import and search follow-up
|
|
90
|
+
|
|
91
|
+
Large input files (at least 1 MiB of input bytes) named `.json`, `.trace`,
|
|
92
|
+
`.cpuprofile` or `.heapprofile`, optionally gzip-compressed, now use a browser
|
|
93
|
+
worker when available. The existing format detector is reused, preserving its
|
|
94
|
+
dispatch order. HTTP/profileURL, local File and CLI base64 loading use this path.
|
|
95
|
+
Input buffers and numeric profile tables transfer ownership instead of being
|
|
96
|
+
JSON-serialized or cloned as full call-tree objects. File inputs can be read in
|
|
97
|
+
the worker. Symbols are demangled there before returning immutable profile data.
|
|
98
|
+
|
|
99
|
+
The UI shows import progress and cancellation. Cancellation terminates the worker;
|
|
100
|
+
new loads and unmounts cancel old imports, and stale results cannot replace newer
|
|
101
|
+
profiles or clear their loading state. The worker is released before UI object
|
|
102
|
+
reconstruction, which yields between bounded batches so it can also be cancelled.
|
|
103
|
+
Worker creation blocked by a host can fall back to the ordinary importer. An
|
|
104
|
+
unexpected worker crash reports an error rather than trying a detached buffer.
|
|
105
|
+
Small files and other filename extensions retain the ordinary importer; that
|
|
106
|
+
synchronous fallback cannot be interrupted during parsing.
|
|
107
|
+
|
|
108
|
+
The worker code is embedded by the build plugin, so both HTTP and offline builds
|
|
109
|
+
can create it without fetching a worker script or WASM from an external host.
|
|
110
|
+
The transport is an internal implementation detail, not a public interchange API.
|
|
111
|
+
|
|
112
|
+
Tests cover numeric/string frame identity, recursion, source locations, formatting,
|
|
113
|
+
aggregate/derived flags, idle ranges, transferred buffers, deep trees, invalid
|
|
114
|
+
references, cancellation, worker cleanup, host fallback and stale-load races.
|
|
115
|
+
The PLC worker import was cancelled interactively while parsing, then completed;
|
|
116
|
+
its exported JSON matched the ordinary importer exactly (four profiles, 17,510
|
|
117
|
+
frames and 1,039,656 call events). A local snapshot smoke check measured 214 ms to
|
|
118
|
+
encode the PLC profiles and 34,602,732 bytes of numeric tables (excluding frame
|
|
119
|
+
strings and metadata). Browser elapsed-time/peak-RSS speedups are not claimed.
|
|
120
|
+
|
|
121
|
+
Search also supports literal whole-word and exact matching. Patterns are reused
|
|
122
|
+
per query and file-only checks avoid allocating name highlight ranges. Unicode
|
|
123
|
+
word boundaries, punctuation, escaped metacharacters and full-name/path matching
|
|
124
|
+
have regressions. User-provided regex is still not enabled.
|
|
125
|
+
|
|
126
|
+
Profile trees and rendering remain in-memory operations. Unsupported formats
|
|
127
|
+
still use the original full-document importer. UI hydration and rendering have
|
|
128
|
+
costs, and transferring work can increase overall elapsed time even while making
|
|
129
|
+
the interface responsive. This does not fix every OOM report. Derived
|
|
130
|
+
transformations other than frame filtering remain synchronous. Frame filtering
|
|
131
|
+
visits grouped paths and yields during traversal, with cancellation; final tree
|
|
132
|
+
sorting remains synchronous. See [filter behavior and limits](PROFILE_FILTERS.md).
|
|
133
|
+
Export retains the final
|
|
134
|
+
Blob bytes and source profile while bounding temporary serialization chunks
|
|
135
|
+
(apart from an individually huge string). An out-of-core model, heap budgets,
|
|
136
|
+
worker-backed transforms, signed comparisons and synchronized timelines remain
|
|
137
|
+
separate follow-ups in [the maintained backlog](ISSUE_BACKLOG.md).
|
|
138
|
+
|
|
139
|
+
## Streaming duration traces
|
|
140
|
+
|
|
141
|
+
Large duration-event `.json`/`.trace` files, optionally gzip-compressed, now use
|
|
142
|
+
incremental parsing inside the worker. A local File is streamed without reading
|
|
143
|
+
its whole contents; transferred ArrayBuffers are consumed in 64 KiB slices.
|
|
144
|
+
Native `DecompressionStream` inflates gzip incrementally. The streaming parser
|
|
145
|
+
([@streamparser/json](https://github.com/juanjoDiaz/streamparser-json), pinned at
|
|
146
|
+
0.0.26) releases completed event objects from its parent array. We retain the
|
|
147
|
+
duration events, naming metadata and first event needed by the existing Hermes
|
|
148
|
+
discriminator, dropping unused event fields and ignored events. Event arguments
|
|
149
|
+
remain intact because they participate in frame identities.
|
|
150
|
+
|
|
151
|
+
This avoids a full inflated byte buffer, a whole JSON string and a complete
|
|
152
|
+
parsed input object. Out-of-order traces still need stable sorting, so retained
|
|
153
|
+
events and constructed trees continue to scale with the capture. This is not an
|
|
154
|
+
out-of-core profile store or a hard bound on total JavaScript/browser memory.
|
|
155
|
+
|
|
156
|
+
The streaming path rejects excessive input complexity before returning a profile:
|
|
157
|
+
|
|
158
|
+
| Limit | Default |
|
|
159
|
+
| --- | ---: |
|
|
160
|
+
| Decompressed input | 1 GiB |
|
|
161
|
+
| Pending event/metadata bytes between completed events | 16 MiB |
|
|
162
|
+
| Retained event source bytes | 256 MiB |
|
|
163
|
+
| Input events, including ignored events | 2,000,000 |
|
|
164
|
+
| JSON nesting depth | 128 |
|
|
165
|
+
|
|
166
|
+
Checks run before bounded parser writes and while tokens/events are emitted.
|
|
167
|
+
The pending-byte check is conservative at chunk boundaries. These are source
|
|
168
|
+
complexity limits, not heap accounting: object overhead, expanded Unicode,
|
|
169
|
+
sorting, call-tree nodes, numeric snapshots, hydration and GPU resources have
|
|
170
|
+
additional costs. A limit failure stops with an actionable error; it never
|
|
171
|
+
silently truncates the trace or retries the full importer. Cancellation releases
|
|
172
|
+
the reader and terminates the worker; source cancellation failures cannot hide
|
|
173
|
+
the original error.
|
|
174
|
+
|
|
175
|
+
Native CPU/heap/speedscope files and known filename overrides retain their
|
|
176
|
+
detector. CPU markers anywhere in the stream, sample tables, duplicate root
|
|
177
|
+
`traceEvents` properties and unrecognized wrapper properties return to that
|
|
178
|
+
detector, preserving their original semantics. Those fallback formats do not
|
|
179
|
+
gain these streaming budgets. Gzip also falls back when native decompression is
|
|
180
|
+
unavailable. URL/CLI loading still buffers the input response before transfer;
|
|
181
|
+
local File streaming avoids that buffer. Tolerated trailing commas and missing
|
|
182
|
+
final `]` after complete array events remain supported. Incomplete values and
|
|
183
|
+
corrupt gzip fail rather than producing a partial result.
|
|
184
|
+
|
|
185
|
+
Validation includes every duration-trace fixture against the ordinary importer,
|
|
186
|
+
one-byte UTF-8/escape boundaries, late metadata, argument identity, unordered and
|
|
187
|
+
anonymous events, sampled/CPU fallback, ignored payloads, gzip File imports with
|
|
188
|
+
whole-file reads forbidden, resource-limit failures, and cancellation races.
|
|
189
|
+
The PLC capture exported identically: four profiles, 17,510 frames, 1,039,656
|
|
190
|
+
call events and matching totals. Fresh-process Windows/Node 24 import-only runs:
|
|
191
|
+
|
|
192
|
+
| Path | Import time | Peak RSS | Heap after import |
|
|
193
|
+
| --- | ---: | ---: | ---: |
|
|
194
|
+
| Ordinary importer | 2,281 ms | 999 MiB | 572 MiB |
|
|
195
|
+
| Streaming duration importer | 4,311 ms | 623 MiB | 437 MiB |
|
|
196
|
+
|
|
197
|
+
This single capture/run measured about 38% lower peak RSS at a latency cost.
|
|
198
|
+
Browser peak memory, rendering and worker hydration are not represented by this
|
|
199
|
+
Node benchmark. Results vary by machine, input and garbage collection. Reproduce
|
|
200
|
+
each path in its own process with `node scripts/benchmark-profile.mjs <capture>
|
|
201
|
+
import` and `node scripts/benchmark-profile.mjs <capture> stream-import`.
|
|
202
|
+
|
|
203
|
+
Issues #519, #347, #478, #386 and #404 remain **partial** in the backlog: the
|
|
204
|
+
specific original reproductions and a general bounded-memory profile model still
|
|
205
|
+
need work.
|
|
206
|
+
|
|
207
|
+
## Faster streaming event parsing
|
|
208
|
+
|
|
209
|
+
CPU profiling of the first streaming implementation showed substantial time in
|
|
210
|
+
token parsing, per-property assignment, selector matching and callbacks. The
|
|
211
|
+
worker now scans structural bytes to frame each complete event object, then
|
|
212
|
+
decodes and parses that one object with native `JSON.parse`. The token parser
|
|
213
|
+
still validates the wrapper/array grammar and handles other values. The bounded
|
|
214
|
+
event framing scan tracks string escapes and nesting; it does not interpret
|
|
215
|
+
fields or split on text delimiters. A rare permissive retry retains the existing
|
|
216
|
+
trailing-comma behavior.
|
|
217
|
+
|
|
218
|
+
No whole-document string or inflated buffer is introduced. Events spanning read
|
|
219
|
+
boundaries use a temporary buffer for that one event. The existing decompressed,
|
|
220
|
+
pending, retained, event-count and depth limits remain in force, including while
|
|
221
|
+
framing native records. Truncated later records cannot be silently dropped when
|
|
222
|
+
repairing a producer's omitted final array bracket. Additional regressions cover
|
|
223
|
+
all small chunk sizes, large records, split UTF-8/escapes, malformed inner and
|
|
224
|
+
outer grammar, and the existing fallback/cancellation/limit cases.
|
|
225
|
+
|
|
226
|
+
Three alternating fresh-process runs of the same PLC capture on Windows/Node 24
|
|
227
|
+
compared the previous streaming bundle (`39b264b`) with native event framing:
|
|
228
|
+
|
|
229
|
+
| Metric | Previous streaming | Native event framing |
|
|
230
|
+
| --- | ---: | ---: |
|
|
231
|
+
| Import times | 4,255 / 4,336 / 4,423 ms | 2,493 / 2,563 / 2,445 ms |
|
|
232
|
+
| Median import time | 4,336 ms | 2,493 ms |
|
|
233
|
+
| Peak RSS | 612 / 613 / 615 MiB | 579 / 573 / 581 MiB |
|
|
234
|
+
| Median peak RSS | 613 MiB | 579 MiB |
|
|
235
|
+
|
|
236
|
+
That is about 43% faster while retaining (and slightly improving) the previous
|
|
237
|
+
memory reduction. These are import-only process measurements, not browser
|
|
238
|
+
rendering or hydration measurements. Each revision's `benchmark-profile.ts` was
|
|
239
|
+
bundled separately with esbuild (Node/CommonJS, ES2020, binary WASM) and run with
|
|
240
|
+
the capture path and `stream-import` action, alternating revisions without
|
|
241
|
+
explicit garbage collection. Captures and timing logs are not committed.
|
|
242
|
+
The full exported profiles match the ordinary importer, including all four
|
|
243
|
+
profiles, 17,510 frames, 1,039,656 call events and profile totals.
|
|
244
|
+
|
|
245
|
+
The same follow-up fixes issue #199's deep-zoom axis labels: formatting receives
|
|
246
|
+
the visible tick interval, tick generation clamps below floating-point precision,
|
|
247
|
+
and overlapping labels are skipped. V8 source-path extraction and exact-first
|
|
248
|
+
sourcemap matching (with empty maps ignored) improve #446, which remains partial
|
|
249
|
+
because automatic linked map discovery/loading is still separate work. The
|
|
250
|
+
maintained backlog records these completion boundaries explicitly.
|
|
251
|
+
|
|
252
|
+
## Smaller duration queues and bounded selected-stack details
|
|
253
|
+
|
|
254
|
+
The 2026-11-20 follow-up reuses complete (`X`) events as begin records rather
|
|
255
|
+
than spreading every event into another full object. Synthesized end records
|
|
256
|
+
retain only their timestamp, name and shared arguments. Source positions and
|
|
257
|
+
argument-based frame identity remain unchanged. Queues already ordered by time
|
|
258
|
+
skip stable-sort workspace; unordered queues still use the existing stable sort,
|
|
259
|
+
including equal-timestamp and epoch-rounding behavior. No additional event index
|
|
260
|
+
is retained. The streaming limits are unchanged.
|
|
261
|
+
|
|
262
|
+
Three alternating fresh-process runs on Windows/Node 24 compared `f6c5d43`
|
|
263
|
+
against this change, using the same ES2020 CommonJS benchmark bundle options:
|
|
264
|
+
|
|
265
|
+
| Capture / metric | Before | After |
|
|
266
|
+
| --- | ---: | ---: |
|
|
267
|
+
| PLC import times (ms) | 2,609 / 2,524 / 2,579 | 2,562 / 2,546 / 2,562 |
|
|
268
|
+
| PLC median peak RSS (MiB) | 576 | 576 |
|
|
269
|
+
| Complete-event fixture import times (ms) | 1,131 / 833 / 822 | 986 / 767 / 779 |
|
|
270
|
+
| Complete-event fixture median peak RSS (MiB) | 362 | 295 |
|
|
271
|
+
| Complete-event fixture median heap at import (MiB) | 224 | 155 |
|
|
272
|
+
|
|
273
|
+
The synthetic fixture contains 250,000 sequential complete events with 5,000
|
|
274
|
+
distinct argument-based frame keys. Its median peak RSS fell about 19%; PLC
|
|
275
|
+
performance remained essentially unchanged. These are process import measurements,
|
|
276
|
+
not browser hydration/rendering or hard heap bounds. Separate full-export checks
|
|
277
|
+
matched the baseline hash for both captures. PLC still exports four profiles,
|
|
278
|
+
17,510 frames and 1,039,656 call events. Captures and benchmark artifacts are not
|
|
279
|
+
committed. Reproduce import measurements with `node scripts/benchmark-profile.mjs
|
|
280
|
+
<capture> stream-import`; do not include export validation in import-only RSS.
|
|
281
|
+
|
|
282
|
+
Selected-stack details now retain/render at most 100 ancestor rows and follow
|
|
283
|
+
parent pointers only as pages are requested, without building the full ancestor
|
|
284
|
+
list. All scopes remain reachable through older-page and reset controls. Clicking
|
|
285
|
+
one selects the exact ancestor instance; see [stack navigation](SEARCH_AND_LAYOUT.md#navigate-selected-stack-ancestors).
|
|
286
|
+
Tests include a 100,000-level chain with a guarded page boundary. The PLC's
|
|
287
|
+
deepest seven-scope stack was paged without changing the export hash.
|
|
288
|
+
|
|
289
|
+
Issues #519, #478, #386, #347 and #404 remain partial: the final profile model,
|
|
290
|
+
worker snapshots, unordered sorts and graph rendering still grow with capture
|
|
291
|
+
size. This work reduces specific temporary allocations and bounds the details
|
|
292
|
+
panel, without claiming an out-of-core viewer or fixing every reported OOM.
|
|
293
|
+
|
|
294
|
+
Validation passed 339 Jest tests across 58 suites (307 snapshots), five Node
|
|
295
|
+
tests, type checking, linting, schema/backlog checks, both builds and a dry-run
|
|
296
|
+
npm contents check. Browser smoke checks used a 220-level profile to verify
|
|
297
|
+
100-row pages, access to the root, reset controls, Enter/Space selection and
|
|
298
|
+
visible ancestor highlighting below the time axis, without console errors.
|
|
299
|
+
|
|
300
|
+
## Bounded graph-search state
|
|
301
|
+
|
|
302
|
+
On 2026-11-21, graph search stopped eagerly caching every source symbol and
|
|
303
|
+
allocating navigation geometry/index entries for every matching call. Per query,
|
|
304
|
+
boolean frame results are capped at 4,096 entries and drawn-label ranges at
|
|
305
|
+
512 entries with at most 64 ranges each. Time Order/Left Heavy navigation keeps
|
|
306
|
+
the first 10,000 matches and checks for one further match to establish
|
|
307
|
+
truncation. The search UI displays a `+` lower bound and a visible refinement
|
|
308
|
+
notice; it never claims the truncated count is complete. Matching/highlighting
|
|
309
|
+
can continue beyond that navigation window.
|
|
310
|
+
|
|
311
|
+
Three alternating fresh-process Windows/Node 24 runs compared `a72d309` with
|
|
312
|
+
bounded search on the PLC's first profile (513,009 graph frames), using a broad
|
|
313
|
+
single-space query:
|
|
314
|
+
|
|
315
|
+
| Search-only metric | Before | After |
|
|
316
|
+
| --- | ---: | ---: |
|
|
317
|
+
| Initialization times (ms) | 96 / 89 / 94 | 3 / 3 / 4 |
|
|
318
|
+
| Retained search heap after explicit GC (MiB) | 119.60 / 119.60 / 119.60 | 0.907 / 0.907 / 0.906 |
|
|
319
|
+
| Retained navigation matches | 513,009 | 10,000 (explicitly limited) |
|
|
320
|
+
|
|
321
|
+
This is a deliberate navigation-limit tradeoff, not equivalent full-result
|
|
322
|
+
enumeration. Separate single-run checks found the same 40 `GroupCalls` matches
|
|
323
|
+
(14 → 22 ms) and zero matches for a nonexistent symbol (14 → 25 ms). Those sparse
|
|
324
|
+
queries retain about half as much search heap, but can be slower because cached
|
|
325
|
+
entries are recomputed. These are Node search measurements after import and graph
|
|
326
|
+
construction, not whole-app memory caps or browser latency guarantees.
|
|
327
|
+
The measured initialization counts results and opens the first ten; it does not
|
|
328
|
+
draw labels or measure browser rendering.
|
|
329
|
+
|
|
330
|
+
The PLC import/export check still reports four profiles, 17,510 source frames
|
|
331
|
+
and 1,039,656 call events, with the unchanged baseline export hash. Captures and
|
|
332
|
+
benchmark artifacts remain private/uncommitted. The maintained benchmark adds
|
|
333
|
+
`node scripts/benchmark-profile.mjs <capture> stream-search [query]` for streaming
|
|
334
|
+
import followed by bounded search of the first profile; its default query is a
|
|
335
|
+
single space. It reports the navigation limit and retained search heap separately
|
|
336
|
+
from input/model memory. Graph creation, worker snapshots, sorting, Sandwich
|
|
337
|
+
row lists and the final model remain proportional to recording size; the OOM
|
|
338
|
+
backlog stays partial.
|
|
339
|
+
|
|
340
|
+
Validation passed all 354 Jest tests across 62 suites (307 snapshots), five Node
|
|
341
|
+
tests, type checking, linting, schema/backlog checks, both builds and package
|
|
342
|
+
contents. Regressions cover cache eviction, dense-symbol range limits, exact
|
|
343
|
+
limit detection, late sparse matches, depth-first order and 25,000-level graphs.
|
|
344
|
+
Browser checks on 20,050 calls verified the visible limit notice, both keyboard
|
|
345
|
+
wrap directions, narrowing to a late result, and the root-label axis fix,
|
|
346
|
+
without console errors.
|
package/docs/OFFLINE.md
ADDED
|
@@ -0,0 +1,58 @@
|
|
|
1
|
+
# Hosted offline use
|
|
2
|
+
|
|
3
|
+
The hosted HTTP build caches its application shell after a successful online
|
|
4
|
+
visit in a service-worker-capable secure browser. Once installation has finished,
|
|
5
|
+
the app's entry page can reopen without a network connection. Import a recording
|
|
6
|
+
from your computer again after reopening; an open recording remains in memory
|
|
7
|
+
when the network disconnects.
|
|
8
|
+
|
|
9
|
+
This caches application code, styles, fonts, icons and WebAssembly decoders only.
|
|
10
|
+
Imported recordings, remote `profileURL` responses, the CLI's local profile
|
|
11
|
+
endpoint, source maps supplied by users, the built-in example recording and
|
|
12
|
+
arbitrary network responses are not saved. A `profileURL` link can open the
|
|
13
|
+
offline app shell, but its remote recording still needs a network connection.
|
|
14
|
+
The built-in example also needs the network. Application caching is not a saved
|
|
15
|
+
recording feature.
|
|
16
|
+
|
|
17
|
+
Use the app's ordinary entry URL, such as `/speedscope/` or
|
|
18
|
+
`/speedscope/index.html`. Hash parameters are supported; query-string entry URLs
|
|
19
|
+
are not intercepted. Neither arbitrary paths within the scope nor cross-origin
|
|
20
|
+
URLs are intercepted. Explicit `no-store` requests and range requests bypass
|
|
21
|
+
the worker as well.
|
|
22
|
+
|
|
23
|
+
## Updates
|
|
24
|
+
|
|
25
|
+
Shell revisions are derived from the generated HTML, immutable asset contents
|
|
26
|
+
and worker policy. An update is installed in the background and waits for all
|
|
27
|
+
tabs using the previous app version to close. Open recordings are not reloaded
|
|
28
|
+
or replaced automatically. Close all app tabs, then reopen the app to adopt a
|
|
29
|
+
waiting update. Offline reopening can continue to use the last installed shell
|
|
30
|
+
if no newer version has been downloaded.
|
|
31
|
+
|
|
32
|
+
Each application scope owns at most three shell generations: the active shell,
|
|
33
|
+
the previous waiting shell and the newest installation/waiting shell. One small
|
|
34
|
+
metadata cache holds only active/waiting cache names. A failed new install keeps
|
|
35
|
+
the active and previously waiting versions usable. Activation removes obsolete
|
|
36
|
+
owned shell caches; caches belonging to other apps/scopes are left untouched.
|
|
37
|
+
There is no runtime cache insertion. Builds reject more than 64 shell files or
|
|
38
|
+
more than 16 MiB of shell assets.
|
|
39
|
+
|
|
40
|
+
Browser storage restrictions, quota pressure or manually clearing site data can
|
|
41
|
+
remove the offline copy. The app continues to work online if registration or
|
|
42
|
+
installation is unavailable. The worker uses the browser's ordinary
|
|
43
|
+
[installation and waiting lifecycle](https://developer.mozilla.org/en-US/docs/Web/API/Service_Worker_API/Using_Service_Workers)
|
|
44
|
+
without `skipWaiting`, `clients.claim` or forced reloads.
|
|
45
|
+
|
|
46
|
+
## Build and local behavior
|
|
47
|
+
|
|
48
|
+
`npm run build:pages` emits the worker and hosted registration script.
|
|
49
|
+
`npm run prepack` retains the existing portable standalone ZIP behavior and
|
|
50
|
+
does not emit/register a service worker. The ordinary development server also
|
|
51
|
+
does not register one. File-protocol pages remain independent of service workers.
|
|
52
|
+
|
|
53
|
+
Localhost root pages are excluded to avoid controlling another local app or CLI
|
|
54
|
+
session. A deliberate hosted-build preview under a subdirectory such as
|
|
55
|
+
`http://127.0.0.1:PORT/app/` can register a worker limited to that directory.
|
|
56
|
+
For browser verification, complete an online visit there, stop the server, then
|
|
57
|
+
reload the same entry URL. Use a dedicated preview origin/port and clear its
|
|
58
|
+
site storage afterward if you no longer need the preview's offline shell.
|
|
@@ -0,0 +1,73 @@
|
|
|
1
|
+
# Filter frames
|
|
2
|
+
|
|
3
|
+
Choose **Profile actions → Filter frames…** to create an aggregate view of your
|
|
4
|
+
code or hide framework frames. The original recording stays in the profile
|
|
5
|
+
selector. Select it to return to the complete profile. Filtering applies to the
|
|
6
|
+
active profile, and you can apply another filter to a derived profile.
|
|
7
|
+
|
|
8
|
+
Choose source file path, function name, or either. Enter one case-insensitive
|
|
9
|
+
literal substring per line. Slashes are interchangeable, so `src/app/` also
|
|
10
|
+
matches Windows paths. Multiple include rules are alternatives (OR); any exclude
|
|
11
|
+
match removes the frame, even if an include rule matches it. Leave include blank
|
|
12
|
+
to retain all frames except exclusions. Empty lines are ignored. File rules do
|
|
13
|
+
not match frames without a source path; use function names for those profiles.
|
|
14
|
+
Names match the displayed frame name, including any arguments that a trace
|
|
15
|
+
importer incorporates into it. There are no regular expressions or glob rules.
|
|
16
|
+
|
|
17
|
+
For example, include `src/` and exclude `node_modules/`. Or leave include empty
|
|
18
|
+
and exclude `vendor/` to remove framework frames while keeping unmapped code.
|
|
19
|
+
Up to 128 non-empty rules and 4,096 combined rule characters are accepted.
|
|
20
|
+
|
|
21
|
+
## What the weights mean
|
|
22
|
+
|
|
23
|
+
This hides individual frames and joins their remaining ancestors/callees. It
|
|
24
|
+
does not discard an entire stack merely because it contains a hidden frame.
|
|
25
|
+
Each remaining stack retains its original weight. If a removed leaf has a kept
|
|
26
|
+
ancestor, its weight becomes self weight on the nearest kept ancestor. If no
|
|
27
|
+
frame remains in a stack, that stack's weight is omitted. Idle/empty-stack weight
|
|
28
|
+
is also omitted. Total/self values and percentages are recomputed from retained
|
|
29
|
+
stacks; they are not percentages of the unfiltered recording.
|
|
30
|
+
|
|
31
|
+
Suppose an 8 MiB allocation profile contains:
|
|
32
|
+
|
|
33
|
+
| Stack | Weight |
|
|
34
|
+
| --- | ---: |
|
|
35
|
+
| framework → render | 2 MiB |
|
|
36
|
+
| framework → render → vendorWork | 2 MiB |
|
|
37
|
+
| framework → helper | 2 MiB |
|
|
38
|
+
| framework | 2 MiB |
|
|
39
|
+
|
|
40
|
+
Including `src/` keeps render and helper, producing 6 MiB total. Render has
|
|
41
|
+
4 MiB total and self: the hidden vendorWork leaf contributes to render's self
|
|
42
|
+
weight. Also excluding helper's file produces a 4 MiB profile containing render
|
|
43
|
+
alone. The original 8 MiB profile remains available.
|
|
44
|
+
|
|
45
|
+
## Derived profile behavior
|
|
46
|
+
|
|
47
|
+
Filtered profiles are explicitly aggregate and derived. Time Order is disabled;
|
|
48
|
+
use Left Heavy, Sandwich or Call Tree. Existing Call Tree mode is preserved when
|
|
49
|
+
applying a filter. Filter rules are appended to the profile name, retaining a
|
|
50
|
+
visible history through further transformations and export/reimport. Exports
|
|
51
|
+
include both the original and filtered profiles, their units and aggregate flags.
|
|
52
|
+
“Combine original profiles” excludes derived profiles to avoid double counting.
|
|
53
|
+
|
|
54
|
+
No remaining frames is reported as an error without appending an empty profile.
|
|
55
|
+
Cancel or Escape discards unfinished work. Pasting rules stays inside the dialog
|
|
56
|
+
and does not invoke clipboard-profile import. Replacing the source recording
|
|
57
|
+
aborts/discards the pending result.
|
|
58
|
+
|
|
59
|
+
The implementation visits grouped call paths instead of every original trace
|
|
60
|
+
event, caches each frame's rule match, and yields between bounded batches of
|
|
61
|
+
traversal/stack work. Traversal and cancellation support a 20,000-frame deep-stack
|
|
62
|
+
regression without recursion. Final tree sorting remains synchronous, and the
|
|
63
|
+
original and derived models both occupy memory; this is not an out-of-core filter.
|
|
64
|
+
A local PLC smoke check filtered the selected profile from 16,231 frames to 40
|
|
65
|
+
in about 0.7 seconds, yielding 45 times. That is a single Node measurement, not a
|
|
66
|
+
general browser latency guarantee.
|
|
67
|
+
|
|
68
|
+
## Memory-profile axes
|
|
69
|
+
|
|
70
|
+
Byte profiles now use power-of-two tick intervals in the main graph and minimap:
|
|
71
|
+
for example 512 KiB steps or whole MiB steps, rendered with the app's existing
|
|
72
|
+
KB/MB/GB labels. Other units retain decimal 1/2/5 spacing. Both axes adapt label
|
|
73
|
+
precision to zoom and skip overlapping labels.
|
package/docs/RELEASING.md
CHANGED
|
@@ -34,9 +34,8 @@ gh workflow run pages.yml --repo node-projects/speedscope --ref main
|
|
|
34
34
|
|
|
35
35
|
## First publication: 2.0.0
|
|
36
36
|
|
|
37
|
-
The first fork version
|
|
38
|
-
|
|
39
|
-
to publish under `@node-projects`, then run:
|
|
37
|
+
The first fork version, **2.0.0**, was published manually by the maintainer on
|
|
38
|
+
2026-11-18. The bootstrap process was:
|
|
40
39
|
|
|
41
40
|
```sh
|
|
42
41
|
npm ci
|
|
@@ -44,16 +43,16 @@ npm test
|
|
|
44
43
|
npm publish --access public
|
|
45
44
|
```
|
|
46
45
|
|
|
47
|
-
`prepack` builds the offline app automatically. This first publication
|
|
48
|
-
require GitHub OIDC.
|
|
46
|
+
`prepack` builds the offline app automatically. This first publication did not
|
|
47
|
+
require GitHub OIDC. The trusted publisher below is now configured.
|
|
49
48
|
Creating a `v2.0.0` GitHub release afterward is safe: the workflow detects the
|
|
50
49
|
existing npm version, attaches its standalone ZIP and skips re-publishing npm.
|
|
51
50
|
For a new version, it attaches the newly built npm tarball as well.
|
|
52
51
|
|
|
53
52
|
## npm trusted publishing after the first publication
|
|
54
53
|
|
|
55
|
-
The
|
|
56
|
-
|
|
54
|
+
The npm trusted publisher was created successfully on 2026-11-18 for
|
|
55
|
+
`@node-projects/speedscope`. It authorizes the following GitHub Actions identity:
|
|
57
56
|
|
|
58
57
|
- Organization/user: `node-projects`
|
|
59
58
|
- Repository: `speedscope`
|
|
@@ -61,6 +60,36 @@ The workflow is prepared for npm trusted publishing (OIDC), with Node.js 24 and
|
|
|
61
60
|
- Environment: leave empty (the npm job does not use an environment)
|
|
62
61
|
- Allow the `npm publish` action.
|
|
63
62
|
|
|
63
|
+
The workflow already uses Node.js 24 and `id-token: write`, so future published
|
|
64
|
+
GitHub releases can authenticate through OIDC without an npm token secret.
|
|
65
|
+
The saved publisher also includes npm's default staging permission; the workflow
|
|
66
|
+
uses direct publication. No GitHub environment restriction is configured because
|
|
67
|
+
the publish job does not declare an environment.
|
|
68
|
+
|
|
69
|
+
To inspect the configuration, an authorized npm maintainer can run:
|
|
70
|
+
|
|
71
|
+
```sh
|
|
72
|
+
npm trust list @node-projects/speedscope --json
|
|
73
|
+
```
|
|
74
|
+
|
|
75
|
+
For recovery/recreation, npm 11.15+ supports the equivalent CLI setup:
|
|
76
|
+
|
|
77
|
+
```sh
|
|
78
|
+
npm trust github @node-projects/speedscope --repository node-projects/speedscope --file release.yml --allow-publish
|
|
79
|
+
```
|
|
80
|
+
|
|
81
|
+
npm requires fresh two-factor authentication for these account operations. The
|
|
82
|
+
setup was completed through the CLI's browser authentication flow; no credentials
|
|
83
|
+
were added to the repository. Recreating an existing publisher is not a routine
|
|
84
|
+
release step.
|
|
85
|
+
|
|
86
|
+
New publisher configurations must complete their first successful automated
|
|
87
|
+
publish within **two days**, otherwise npm expires them. Creating a GitHub release
|
|
88
|
+
for an already-published version does not validate OIDC because this workflow
|
|
89
|
+
skips the npm publish step for that version. Use the next new version (for example
|
|
90
|
+
`v2.0.1`) when ready. Until that first workflow publish, setup is verified but an
|
|
91
|
+
end-to-end OIDC publication has not been exercised.
|
|
92
|
+
|
|
64
93
|
See [npm's trusted publisher setup](https://docs.npmjs.com/trusted-publishers/).
|
|
65
94
|
The repository URL in package.json points to this fork, as required for provenance.
|
|
66
95
|
|
|
@@ -69,8 +98,8 @@ only be completed after the package exists and by an authorized npm maintainer.
|
|
|
69
98
|
Keep the workflow filename and repository settings exact. Allow direct `npm publish`
|
|
70
99
|
for this publisher. There is no npm credential included in this repository.
|
|
71
100
|
|
|
72
|
-
|
|
73
|
-
|
|
101
|
+
Version 2.0.0 is available from npm, and the README's npm badge and installation
|
|
102
|
+
command now refer to the published scoped package.
|
|
74
103
|
|
|
75
104
|
## Publish a release
|
|
76
105
|
|