ah-fs.processor 
Processes ah-fs data obtained from async resources related to file system opearations.
Installation
npm install ah-fs.processor
Table of Contents generated with DocToc
processFileSystem
Parameters
$0
Object
$0.activities
Map<string, Object> a map of async activities hashed by id$0.includeActivities
boolean? if true
the actual activities are appended to the output (optional, default false
)
ReadFileProcessor
Instantiates an fs.readFile data processor to process data collected via
nodesource/ah-fs
Parameters
$0
ObjectincludeActivities
boolean? if true
the actual activities are appended to the output (optional, default false
)
readFileProcessor.process
Processes the supplied async activities and splits them into
groups, and operations each representing a file read fs.readFile
.
Groups
The returned value has a groups
property which just lists the ids
of async resources that were grouped together to form an operation
indexed by the id of the open
resource.
Thus the groups
is a map of sets.
If no file read was encountered the groups are empty.
Operations
Additionally an operations
property is included as well. Each operation
represents one full fs.readFile
execution. There will be one operation per
group and they are indexed by the corresponding open resource id
as well.
An operation
has the following properties:
fs.readFile
specific Operation Properties
Data about the async resources that were part of the operation, by default
only id
and triggerId
are included:
- open: contains data about opening the file
- stat: contains data about getting file stats
- read: contains data about reading the file
- close: contains data about closing the file
General Operation Properties
The information below is the same for all operation
s and thus is only
mentioned here and linked from the documentation of all other processors.
Data about the lifetime of the operation:
-
lifeCycle: contains three timestamps that detail when an operation was created,
for how long it was alive and when it was destroyed.
- created: the timestamp when the first resource of the operation was created
- destroyed: the timestamp when the last resource of the operation was destroyed
- timeAlive: the difference between the
destroyed
and created
timestamps, i.e.
how long the operation's resources were alive
Each timestamp has the following two properties provided by utils.prettyNs.
- ns: time in nanoseconds {Number}
- ms: pretty printed time in milliseconds {String}
Data that links to user code that is responsible for the operation occurring.
-
createdAt: provides the line of code that called fs.readFile
-
userFunctions: depending on the settings (see constructor docs) each resource
will include it's own array of userFunctions or they are separated out into
one property with duplicates merged. The latter is the default behavior.
In either case userFunctions
is an Array of Objects with the following properties:
- name: the function name
- inferredName: the inferred function name, only needed if the
name
is not set - file: the file in which the function was defined
- line: the line on which the functino was defined in that file
- column: the column on which the functino was defined in that file
- location: the file and line + column where the function was defined combined into a string
- args: the
err
and information about the res
of the operation
with which the function was invoked - propertyPaths: the object paths at which the function was found, these could be multiple
since the function could've been attached to multiple resources (only available if the functions
were separated from the resources and merged)
- propertyPath: the object path at which the function was found (only available if the
functions weren't separated and thus are still part of each resource)
Sample Return Value
The sample return value was created with default options.
{ groups: Map { 10 => Set { 10, 11, 12, 13 } },
operations:
Map {
10 => { lifeCycle:
{ created: { ms: '44.12ms', ns: 44119000 },
destroyed: { ms: '85.95ms', ns: 85955000 },
timeAlive: { ms: '41.84ms', ns: 41836000 } },
createdAt: 'at Test.<anonymous> (/Volumes/d/dev/js/async-hooks/ah-fs/test/read-one-file.js:36:6)',
open: { id: 10, triggerId: 1 },
stat: { id: 11, triggerId: 10 },
read: { id: 12, triggerId: 11 },
close: { id: 13, triggerId: 12 },
userFunctions:
[ { file: '/Volumes/d/dev/js/async-hooks/ah-fs/test/read-one-file.js',
line: 39,
column: 17,
inferredName: '',
name: 'onread',
location: 'onread (/Volumes/d/dev/js/async-hooks/ah-fs/test/read-one-file.js:39:17)',
args:
{ '0': null,
'1':
{ type: 'Buffer',
len: 6108,
included: 18,
val:
{ utf8: 'const test = requi',
hex: '636f6e73742074657374203d207265717569' } },
proto: 'Object' },
propertyPaths:
[ 'open.resource.context.callback',
'stat.resource.context.callback',
'read.resource.context.callback',
'close.resource.context.callback' ] } ] } } }
Returns Object information about fs.readFile
operations with the
structure outlined above
ReadFileProcessor.operationSteps
The minimum number of steps, represented as an async resource each,
involved to execute fs.readFile
.
This can be used by higher level processors to group
activities looking for larger operations first and then
operations involving less steps.
Steps are: open, stat, read+, close
ReadFileProcessor.operation
Description of the operation: 'fs.readFile'.
ReadFileOperation
Processes a group of async activities that represent a fs read stream operation.
It is used by the ReadFileProcessor as part of process
.
Four operation steps are derived from the group, each providing some information
about the operation in question.
Each step is processed into an operation in the corresponding private method, i.e. _processOpen
.
These methods are documented below for information's sake, they should not be called directly,
nor should you have a need to directly instantiate a ReadFileOperation
in the first place.
Parameters
group
Map<Number, Set<Number>> the ids of the activities that were part of the operationincludeActivities
Boolean? if true
the activities are attached to
each operation step (optional, default false
)
readFileOperation._processOpen
The open resource tells us where in user code the fs.readFile
originated
via the second frame of the stack trace, as well as when the operation
was created.
Additionally it has the same user functions attached as all the other resources.
Parameters
info
Object information about the open step, pre-processed by the ReadFileProcessor
.
readFileOperation._processStat
The stat resource gives us no interesting information.
Therefore we just capture the id
, triggerId
and userFunctions
and if so desired
attach the activities.
Parameters
info
Object information about the open step, pre-processed by the ReadFileProcessor
.
readFileOperation._processRead
The read resource gives us no interesting information.
Therefore we just capture the id
, triggerId
and userFunctions
and if so desired
attach the activities.
Parameters
info
Object information about the read step, pre-processed by the ReadFileProcessor
.
readFileOperation._processClose
The main information we pull from the close resource is the destroy
timestamp.
Combined with the init
timestamp of the open resource it allows us to deduce how long
the file read took.
Parameters
info
Object information about the close step, pre-processed by the ReadFileProcessor
.
readFileOperation.summary
Returns the summary of processing the group into an operation.
The summary of all operations has a very similar structure, but includes some properties that are specific to this
particular operation.
The general properties lifeCycle
and createdAt
are documented as part of
the ReadFileProcessor
.
Therefore learn more here.
Properties Specific to fs.readFile
- open: see
readFileOperation._processOpen
- stat: see
readFileOperation._processStat
- read: see
readFileOperation._processRead
- close: see
readFileOperation._processClose
Parameters
$0
Object options
$0.separateFunctions
Boolean? when true
the user functions are separated out
from the specific operations and attached as a userFunctions
array directly to the returned
result (optional, default true
)$0.mergeFunctions
Boolean? if true
when a duplicate function is found in the
separated functions Array, they are merged into one while preserving all information
from both version. Note that this setting only activates if separateFunctions
is true
as well. (optional, default true
)
Returns Object all important information about the current operation
ReadStreamProcessor
Instantiates an fs.createReadStream data processor to process data collected via
nodesource/ah-fs
Parameters
$0
ObjectincludeActivities
boolean? if true
the actual activities are appended to the output (optional, default false
)
readStreamProcessor.process
Processes the supplied async activities and splits them into
groups, and operations each representing a file read stream fs.createReadStream
.
Groups
The returned value has a groups
property which just lists the ids
of async resources that were grouped together to form an operation
indexed by the fd
on which the readFile operated.
Thus the groups
is a map of sets.
If no file read stream was encountered the groups are empty.
Operations
Additionally an operations
property is included as well. Each operation
represents one full fs.createReadStream
execution. There will be one operation per
group and they are indexed by the corresponding fd
as well.
An operation
has the following properties:
fs.createReadStream
specific Operation Properties
Data about the async resources that were part of the operation, by default
only id
and triggerId
are included:
- open: contains data about opening the file
- stream: contains data about how the stream was configured, including readable state and
the path to the file being read, pipes count, encoding, etc.
- reads: an Array of reads, each containing data about reading a chunk from the file including
the time spent to complete reading the particular chunk
- close: contains data about closing the file
General Operation Properties
Sample Return Value
The sample return value was created with default options.
{ groups: Map { 10 => Set { 10, 12, 13, 14, 16 } },
operations:
Map {
10 => { lifeCycle:
{ created: { ms: '1.60ms', ns: 1600000 },
destroyed: { ms: '14.33ms', ns: 14329000 },
timeAlive: { ms: '12.73ms', ns: 12729000 } },
createdAt: 'at Test.<anonymous> (/Volumes/d/dev/js/async-hooks/ah-fs/test/readstream-one-file.js:94:6)',
open: { id: 10, triggerId: 3 },
stream:
{ id: 14,
triggerId: 12,
path: '/Volumes/d/dev/js/async-hooks/ah-fs/test/readstream-one-file.js',
flags: 'r',
fd: 19,
objectMode: false,
highWaterMark: 65536,
pipesCount: 0,
defaultEncoding: 'utf8',
encoding: null },
reads:
[ { id: 12,
triggerId: 10,
timeSpent: { ms: '0.83ms', ns: 830000 } },
{ id: 13,
triggerId: 12,
timeSpent: { ms: '0.24ms', ns: 240000 } } ],
close: { id: 16, triggerId: 13 },
userFunctions:
[ { file: '/Volumes/d/dev/js/async-hooks/ah-fs/test/readstream-one-file.js',
line: 99,
column: 16,
inferredName: '',
name: 'onend',
location: 'onend (/Volumes/d/dev/js/async-hooks/ah-fs/test/readstream-one-file.js:99:16)',
args: null,
propertyPaths: [ 'stream.resource.args[0]._events.end[1]' ] },
{ file: '/Volumes/d/dev/js/async-hooks/ah-fs/test/readstream-one-file.js',
line: 98,
column: 17,
inferredName: '',
name: 'ondata',
location: 'ondata (/Volumes/d/dev/js/async-hooks/ah-fs/test/readstream-one-file.js:98:17)',
args: null,
propertyPaths: [ 'stream.resource.args[0]._events.data' ] } ] } } }
Returns Object information about fs.createReadStream
operations with the
structure outlined above
ReadStreamProcessor.operationSteps
The minimum number of steps, represented as an async resource each,
involved to execute fs.createReadStream
.
This can be used by higher level processors to group
activities looking for larger operations first and then
operations involving less steps.
Steps are: open, stream+, read+, close
ReadStreamProcessor.operation.null
Description of the operation: 'fs.createReadStream'.
ReadStreamOperation
Processes a group of async activities that represent a fs read stream operation.
It is used by the ReadStreamProcessor as part of process
.
Four operation steps are derived from the group, each providing some information
about the operation in question.
Each step is processed into an operation in the corresponding private method, i.e. _processOpen
.
These methods are documented below for information's sake, they should not be called directly,
nor should you have a need to directly instantiate a ReadStreamOperation
in the first place.
Parameters
group
Map<Number, Set<Number>> the ids of the activities that were part of the operationincludeActivities
Boolean? if true
the activities are attached to
each operation step (optional, default false
)
readStreamOperation._processOpen
An open doesn't have too much info, but we can glean two very important
data points:
- the init timestamp tells us when the stream was created
- the last frame of the init stack tells us where
createReadStream
was called.
Parameters
info
Object information about the open step, pre-processed by the ReadStreamProcessor
.
readStreamOperation._processTick
The ReadStream Tick gives us a lot of information. It has an args array with
the ReadStream and its ReadableState respectively
The ReadStream provides us the following:
- the path to the file we are streaming
- the flags with which the file was opened
- the fd (assuming we are dealing with the tick triggered indirectly by the open)
All callbacks on the _events of the ReadStream have been removed, but are present
inside the functions object (see below).
The ReadableState provides us the following:
- objectMode
true|false
- highWaterMark
- pipesCount
- defaultEncoding, i.e. utf8
- encoding, i.e. utf8
The information extracted from the tick is attached to a stream
property
provided with the summary
.
Parameters
info
Object information about the tick step, pre-processed by the ReadStreamProcessor
.
readStreamOperation._processRead
The read resource doesn't give us too much information.
The stack traces originate in core and we don't see any registred
user callbacks, as those are present on the stream instead.
However we can count the amount of reads that occurred and deduce how
long each read took from the before
and after
timestamps.
Parameters
info
Object information about the read step, pre-processed by the ReadStreamProcessor
.
readStreamOperation._processClose
The main information we pull from the close resource is the destroy
timestamp.
Combined with the init
timestamp of the open resource it allows us to deduce how long
the read stream was active.
Parameters
info
Object information about the close step, pre-processed by the ReadStreamProcessor
.
readStreamOperation.summary
Returns the summary of processing the group into an operation.
The summary of all operations has a very similar structure, but includes some properties that are specific to this
particular operation.
The general properties lifeCycle
and createdAt
are documented as part of
the ReadFileProcessor
.
Therefore learn more here.
Properties Specific to fs.createReadStream
- open: see
readStreamOperation._processOpen
- stream: see
readStreamOperation._processTick
- read: see
readStreamOperation._processRead
- close: see
readStreamOperation._processClose
Parameters
$0
Object options
$0.separateFunctions
Boolean? when true
the user functions are separated out
from the specific operations and attached as a userFunctions
array directly to the returned
result (optional, default true
)$0.mergeFunctions
Boolean? if true
when a duplicate function is found in the
separated functions Array, they are merged into one while preserving all information
from both version. Note that this setting only activates if separateFunctions
is true
as well. (optional, default true
)
Returns Object all important information about the current operation
openInitFrame0Rx
Sample initStack of writeFile open, calles as first operation of fs.writeFile
.
In order to be sure this is a writeFile open we need to check the two topmost frames.
"at Object.fs.open (fs.js:581:11)",
"at Object.fs.writeFile (fs.js:1155:6)",
"at Test. (/Volumes/d/dev/js/async-hooks/ah-fs/test/write-one-file.js:28:6)",
Code at fs.js:581:
binding.open(pathModule._makeLong(path), ...
Code at fs.js:1155:
fs.open(path, flag, options.mode, function(openErr, fd) ...
Bottom frame has info about where the call fs.writeFile
originated.
writeInitFrame0Rx
Sample init stack of writeFile write, called afer fs.open
completes:
"at Object.fs.write (fs.js:643:20)",
"at writeAll (fs.js:1117:6)",
"at writeFd (fs.js:1168:5)",
"at fs.js:1159:7",
"at FSReqWrap.oncomplete (fs.js:117:15)"
Code at fs.js:643:
binding.writeBuffer(fd, buffer, offset, length, position, req);
closeInitFrame0Rx
Sample initStack of writeFile close, called after last fs.write
completes:
"at Object.fs.close (fs.js:555:11)",
"at fs.js:1131:14",
"at FSReqWrap.wrapper as oncomplete"
Code at fs.js:555:
binding.close(fd, req);
WriteFileProcessor
Instantiates an fs.writeFile data processor to process data collected via
nodesource/ah-fs
Parameters
$0
ObjectincludeActivities
boolean? if true
the actual activities are appended to the output (optional, default false
)
writeFileProcessor.process
Processes the supplied async activities and splits them into
groups, and operations each representing a file read stream fs.createWriteFile
.
Groups
The returned value has a groups
property which just lists the ids
of async resources that were grouped together to form an operation
indexed by the id of the fs.open
activity that was part of the fs.writeFile
.
Thus the groups
is a map of sets.
If no file write file was encountered the groups are empty.
Operations
Additionally an operations
property is included as well. Each operation
represents one full fs.writeFile
execution. There will be one operation per
group and they are indexed by the corresponding open id as well.
An operation
has the following properties:
fs.createWriteFile
specific Operation Properties
Data about the async resources that were part of the operation, by default
only id
and triggerId
are included:
- open: contains data about opening the file
- writes: an Array of writes, each containing data about writing a
chunk from the file including the time spent to complete writing the
particular chunk
- close: contains data about closing the file
General Operation Properties
Sample Return Value
The sample return value was created with default options.
{ groups: Map { 10 => Set { 10, 11, 12 } },
operations:
Map {
10 => { lifeCycle:
{ created: { ms: '24.49ms', ns: 24491000 },
destroyed: { ms: '33.96ms', ns: 33964000 },
timeAlive: { ms: '9.47ms', ns: 9473000 } },
createdAt: 'at Test.<anonymous> (/Volumes/d/dev/js/async-hooks/ah-fs/test/write-one-file.js:28:6)',
open: { id: 10, triggerId: 1 },
write: { id: 11, triggerId: 10 },
close: { id: 12, triggerId: 11 } } } }
Returns Object information about fs.createWriteFile
operations with the
structure outlined above
WriteFileProcessor.operationSteps
The minimum number of steps, represented as an async resource each,
involved to execute fs.writeFile
.
This can be used by higher level processors to group
activities looking for larger operations first and then
operations involving less steps.
Steps are: open, write+, close
WriteFileProcessor.operation.null
Description of the operation: 'fs.writeFile'.
WriteFileOperation
Processes a group of async activities that represent a fs write file operation.
It is used by the WriteFileProcessor as part of process
.
Three operation steps are derived from the group, each providing some information
about the operation in question.
Each step is processed into an operation in the corresponding private method, i.e. _processOpen
.
These methods are documented below for information's sake, they should not be called directly,
nor should you have a need to directly instantiate a WriteFileOperation
in the first place.
Parameters
group
Map<Number, Set<Number>> the ids of the activities that were part of the operationincludeActivities
Boolean? if true
the activities are attached to
each operation step (optional, default false
)
writeFileOperation._processOpen
The open resource tells us where in user code the fs.writeFile
originated
via the second frame of the stack trace, as well as when the operation
was created.
Parameters
info
Object information about the open step, pre-processed by the WriteFileProcessor
.
writeFileOperation._processWrite
The write resource gives us no interesting information.
Therefore we just capture the id
, triggerId
and if so desired
attach the activities.
Parameters
info
Object information about the write step, pre-processed by the WriteFileProcessor
.
writeFileOperation._processClose
The main information we pull from the close resource is the destroy
timestamp.
Combined with the init
timestamp of the open resource it allows us to deduce how long
the file write took.
Parameters
info
Object information about the close step, pre-processed by the WriteFileProcessor
.
writeFileOperation.summary
Returns the summary of processing the group into an operation.
The summary of all operations has a very similar structure, but includes some properties that are specific to this
particular operation.
The general properties lifeCycle
and createdAt
are documented as part of
the WriteFileProcessor
.
Therefore learn more here.
Properties Specific to fs.writeFile
- open: see
writeFileOperation._processOpen
- write: see
writeFileOperation._processWrite
- close: see
writeFileOperation._processClose
Note this summary function takes no parameters (like the other Operations) since
we don't find any user functions related to the write file operation and thus
have nothing to process.
Returns Object all important information about the current operation
WriteStreamProcessor
Instantiates an fs.createWriteStream data processor to process data collected via
nodesource/ah-fs
Parameters
$0
Object
$0.includeActivities
boolean? if true
the actual activities are appended to the output (optional, default false
)$0.separateFunctions
Boolean? when true
the user functions are separated out
from the specific resources and attached as a userFunctions
array directly to the returned
operations (optional, default true
)
writeStreamProcessor.process
Processes the supplied async activities and splits them into
groups, and operations each representing a file write stream fs.createWriteStream
.
Groups
The returned value has a groups
property which just lists the ids
of async resources that were grouped together to form an operation
indexed by the fd
on which the writeFile operated.
Thus the groups
is a map of sets.
If no file write stream was encountered the groups are empty.
Operations
Additionally an operations
property is included as well. Each operation
represents one full fs.createWriteStream
execution. There will be one operation per
group and they are indexed by the corresponding fd
as well.
An operation
has the following properties:
fs.createWriteStream
specific Operation Properties
Data about the async resources that were part of the operation, by default
only id
and triggerId
are included:
- open: contains data about opening the file
- stream: contains data about how the stream was configured, including writeable state and
the path to the file being write, pipes count, encoding, etc.
- writes: an Array of writes, each containing data about writing a chunk from the file including
the time spent to complete writeing the particular chunk
- close: contains data about closing the file
General Operation Properties
Sample Return Value
The sample return value was created with default options.
{ groups: Map { 10 => Set { 14, 10, 16, 19 } },
operations:
Map {
10 => { lifeCycle:
{ created: { ms: '1.12ms', ns: 1123000 },
destroyed: { ms: '18.20ms', ns: 18205000 },
timeAlive: { ms: '17.08ms', ns: 17082000 } },
createdAt: 'at Test.<anonymous> (/Volumes/d/dev/js/async-hooks/ah-fs/test/read-stream-piped-into-write-stream.js:29:26)',
open: { id: 10, triggerId: 3 },
stream:
{ id: 16,
triggerId: 13,
path: '/dev/null',
flags: 'w',
fd: 19,
mode: 438 },
writes:
[ { id: 14,
triggerId: 13,
timeSpent: { ms: '0.14ms', ns: 139000 } } ],
close: { id: 19, triggerId: 15 },
userFunctions:
[ { file: '/Volumes/d/dev/js/async-hooks/ah-fs/test/read-stream-piped-into-write-stream.js',
line: 32,
column: 19,
inferredName: '',
name: 'onfinish',
location: 'onfinish (/Volumes/d/dev/js/async-hooks/ah-fs/test/read-stream-piped-into-write-stream.js:32:19)',
args: null,
propertyPaths: [ 'stream.resource.args[1].pipes._events.finish[1]' ] } ] } } }
Returns Object information about fs.createWriteStream
operations with the
structure outlined above
writeStreamProcessor._separteIntoGroups
Here we try our best to piece together the parts of a WriteStream,
Open | Write+ | WriteStreamTick | Close.
Since they aren't linked by a common file descriptor or similar we rely
on async resource graph structure and the timestamps to take a best guess.
We just don't have the data available to piece this together with 100% certainty.
Below is a sample of collected async resources with all but types and ids removed.
{ type: 'FSREQWRAP', id: 10, tid: 3 } open, write stream triggered by root
{ type: 'FSREQWRAP', id: 11, tid: 3 } open, read stream triggered by root
{ type: 'TickObject', id: 12, tid: 3 } read stream tick, triggered by root
{ type: 'FSREQWRAP', id: 13, tid: 11 } read, triggerd by open of read steam
{ type: 'FSREQWRAP', id: 14, tid: 13 } write, triggerd by read of read steam
{ type: 'FSREQWRAP', id: 15, tid: 13 } read, next chunk, triggered by first read
{ type: 'TickObject', id: 16, tid: 13 } stream tick, triggerd by first read
{ type: 'FSREQWRAP', id: 18, tid: 15 } close read stream, triggered by last read
{ type: 'FSREQWRAP', id: 19, tid: 15 } close write stream, triggered by last read
We reason about that data as follows in order to piece together the WriteStream.
Connecting WriteSteam Write to WriteStream Close
Write (id: 14) is triggered by read of read stream (id: 13).
The same read triggers the last read (id: 15).
That last read triggers the close of the write stream (id: 19).
Therefore we can connect the write stream write to the write stream close since they
have a common parent in their ancestry (the first read of the read stream).
-- Read2:15 -- WriteStream:Close:19
/
Read1:13
\
-- WriteStream:Write:14
However I would imagine that this breaks down once we have on read stream piped into multiple
write streams as then the writes have the same Read parent.
Connecting WriteStream Open to WriteStream Write
There is no 100% way to get this right, but if we assume that the first write happens right after
the opening of the write stream in the same context we can do the following.
We already know that the common parent of WriteStream:Write and
WriteStream:Close is Read1:13.
Therefore we find all WriteStream:Opens that share a parent with Read1:13.
The ones with the closest parent win.
If we find more than one, we pick the one that was initialized closest to
the WriteStream:Write timewise, assuming that we write to the stream
immediately after opening it.
-- ReadStream:Open:11 -- Read1:13 -- Read2:15 -- WriteStream:Close:19
/ \
Parent:3 -- WriteStream:Write:14
\
-- WriteStream:Open:10
WriteStreamProcessor.operationSteps
The minimum number of steps, represented as an async resource each,
involved to execute fs.createWriteStream
.
This can be used by higher level processors to group
activities looking for larger operations first and then
operations involving less steps.
Steps are: open, stream, write+, close
WriteStreamProcessor.operation.null
Description of the operation: 'fs.createWriteStream'.
WriteStreamOperation
Processes a group of async activities that represent a fs write stream operation.
It is used by the writeStreamProcessor as part of process
.
Four operation steps are derived from the group, each providing some information
about the operation in question.
Each step is processed into an operation in the corresponding private method, i.e. _processOpen
.
These methods are documented below for information's sake, they should not be called directly,
nor should you have a need to directly instantiate a writeStreamOperation
in the first place.
Parameters
group
Map<Number, Set<Number>> the ids of the activities that were part of the operationincludeActivities
Boolean? if true
the activities are attached to
each operation step (optional, default false
)
writeStreamOperation._processOpen
An open doesn't have too much info, but we can glean two very important
data points:
- the init timestamp tells us when the stream was created
- the last frame of the init stack tells us where
createWriteStream
was called.
Parameters
info
Object information about the open step, pre-processed by the WriteStreamProcessor
.
writeStreamOperation._processTick
The WriteStream Tick gives us a lot of information. It is the same tick object that we process
in the ReadStreamOperation to glean data about the read stream.
It has an args array with the ReadStream and its ReadableState respectively.
The ReadableState included the WritableState which the ah-fs pre-processor
already plucked for us and added as the 3rd argument.
Additionally it includes lots of functions including user functions registered with the
WriteStream, i.e. on('finish')
.
Ergo the WriteStream provides us the following as part of the WritableState:
- the path to the file we are writing into
- the flags with which the file was opened
- the fd (assuming we are dealing with the tick triggered indirectly by the open)
All callbacks on the _events of the ReadStream and WriteStream have been removed, but are present
inside the functions object (see below).
The information extracted from the tick is attached to a stream
property
provided with the summary
.
Parameters
info
Object information about the tick step, pre-processed by the WriteStreamProcessor
.
writeStreamOperation._processwrite
The write resource doesn't give us too much information.
The stack traces originate in core and we don't see any registred
user callbacks, as those are present on the stream instead.
However we can count the amount of writes that occurred and deduce how
long each write took from the before
and after
timestamps.
Parameters
info
Object information about the write step, pre-processed by the WriteStreamProcessor
.
writeStreamOperation._processClose
The main information we pull from the close resource is the destroy
timestamp.
Combined with the init
timestamp of the open resource it allows us to deduce how long
the write stream was active.
Parameters
info
Object information about the close step, pre-processed by the WriteStreamProcessor
.
License
MIT