Massive Heap Snapshot Parser
Install
To install just do:
$ npm install mhsp
Usage
This API will extract "snapshot"
, "nodes"
, "edges"
and "strings"
from
very large snapshots in a way that allows you to continue working with them. It
doesn't however show "trace_function_infos"
, "trace_tree"
or "samples"
.
Mainly because I don't need them right now. File an issue or open a PR if you'd
like to help.
parseSnapshot(path)
- Returns
DataAccessor
instance
This does all the magic. Just give it a path to a snapshot and it'll automatically do everything for you. Here's an example:
'use strict';const parseSnapshot = ; const accessor = ;
Warning that for large snapshots this can take several minutes.
accessor.snapshot
- Is
Object
This is the metadata object that contains various information about the snapshot.
accessor.nodes
- Is a
Uint32Array
This is a Uint32Array
of all the nodes in the snapshot. Go ahead and access
it by index.
accessor.edges
- Is a
Uint32Array
This is also a Uint32Array
that can be accessed by index.
accessor.getString(index)
- Returns
String
Returning strings from the "strings"
field isn't as straight forward. Some
fun index tracking is done under the hood so the entire "strings"
section can
live in one big Buffer
.
accessor.writeToFile(path)
- Returns
Number
of bytes written to disk
Since generating the accessor takes so long you can go ahead and write the
entire thing to disk in the form of a binary blob. Can then use importBin()
to read it back in later. Is much much faster.
importBin(path)
- Returns
DataAccessor
instance
Read in files that have already been processed and written to disk as binary blobs. On my i7 the first time I process a 2GB snapshot can take over 1.5 minutes. But reading it back in this way only takes a few seconds. Highly recommended.