nodejs
/
node
mirror of https://github.com/nodejs/node.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

doc: add error documentation 

PR-URL: https://github.com/iojs/io.js/pull/789
Reviewed-by: Rod Vagg <rod@vagg.org>
Reviewed-By: Trevor Norris <trev.norris@gmail.com>
Reviewed-By: Jeremiah Senkpiel <fishrock123@rocketmail.com>
This commit is contained in:
Chris Dickinson 2015-02-10 12:48:55 -08:00
parent d832be4aa3
commit 7e2235aebb
3 changed files with 472 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
doc/api/_toc.markdown
View File

@ -12,6 +12,7 @@
* [Debugger](debugger.html)
* [DNS](dns.html)
* [Domain](domain.html)
* [Errors](errors.html)
* [Events](events.html)
* [File System](fs.html)
* [Globals](globals.html)

1
doc/api/all.markdown
View File

@ -11,6 +11,7 @@
@include dgram
@include dns
@include domain
@include errors
@include events
@include fs
@include globals

470
doc/api/errors.markdown Normal file
View File

@ -0,0 +1,470 @@
# Errors
<!--type=misc-->
Errors generated by io.js fall into two categories: JavaScript errors and system
errors. All errors inherit from or are instances of JavaScript's [Error](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Error)
class and are guaranteed to provide *at least* the attributes available on that
class.
When an operation is not permitted due to language-syntax or
language-runtime-level reasons, a **JavaScript error** is generated and thrown
as an **exception**. If an operation is not allowed due to system-level
restrictions, a **system error** is generated. Client code is then given the
opportunity to **intercept** this error based on how the API **propagates** it.
The style of API called determines how generated errors are handed back, or
**propagated**, to client code, which in turn informs how the client may **intercept**
the error. Exceptions can be intercepted using the `try / catch` construct;
other propagation strategies are covered [below](#errors_error_propagation_and_interception).
## JavaScript Errors
<!--type=misc-->
JavaScript errors typically denote that an API is being used incorrectly, or that
there is a problem with the program as written.
### Class: Error
<!--type=class-->
A general error object. Unlike other error objects, `Error` instances do not
denote any specific circumstance of why the error occurred. Errors capture a
"stack trace" detailing the point in the program at which they were
instantiated, and may provide a description of the error.
**Note**: io.js will generate this class of error to encapsulate system
errors as well as plain JavaScript errors.
#### new Error(message)
Instantiates a new Error object and sets its `.message` property to the provided
message. Its `.stack` will represent the point in the program at which `new Error`
was called. Stack traces are subject to [V8's stack trace API](https://code.google.com/p/v8-wiki/wiki/JavaScriptStackTraceApi).
Stack traces only extend to the beginning of synchronous code execution, *or* a number of frames given by
`Error.stackTraceLimit`, whichever is smaller.
#### error.message
A string of the value passed to `Error()` upon instantiation. The message will
also appear in the first line of the stack trace of the error. Changing this
property *may not* change the first line of the stack trace.
#### error.stack
A property that, when **accessed**, returns a string representing the point in the program
at which this error was instantiated. An example stacktrace follows:
Error: Things keep happening!
at /home/gbusey/file.js:525:2
at Frobnicator.refrobulate (/home/gbusey/business-logic.js:424:21)
at Actor.<anonymous> (/home/gbusey/actors.js:400:8)
at increaseSynergy (/home/gbusey/actors.js:701:6)
The first line is formatted as `<error class name>: <error message>`, and it is followed
by a series of stack frames (each line beginning with "at "). Each frame describes
a call site in the program that lead to the error being generated. V8 attempts to
display a name for each function (by variable name, function name, or object
method name), but occasionally it will not be able to find a suitable name. If
V8 cannot determine a name for the function, only location information will be
displayed for that frame. Otherwise, the determined function name will be displayed
with location information appended in parentheses.
Frames are **only** generated for JavaScript functions. If, for example, execution
synchronously passes through a C++ addon function called `cheetahify`, which itself
calls a JavaScript function, the frame representing the `cheetahify` call will **not**
be present in stacktraces:
```javascript
var cheetahify = require('./native-binding.node');
function makeFaster() {
// cheetahify *synchronously* calls speedy.
cheetahify(function speedy() {
throw new Error('oh no!');
});
}
makeFaster(); // will throw:
// /home/gbusey/file.js:6
// throw new Error('oh no!');
// ^
// Error: oh no!
// at speedy (/home/gbusey/file.js:6:11)
// at makeFaster (/home/gbusey/file.js:5:3)
// at Object.<anonymous> (/home/gbusey/file.js:10:1)
// at Module._compile (module.js:456:26)
// at Object.Module._extensions..js (module.js:474:10)
// at Module.load (module.js:356:32)
// at Function.Module._load (module.js:312:12)
// at Function.Module.runMain (module.js:497:10)
// at startup (node.js:119:16)
// at node.js:906:3
```
The location information will be one of:
* `native`, if the frame represents a call internal to V8 (as in `[].forEach`).
* `plain-filename.js:line:column`, if the frame represents a call internal to io.js.
* `/absolute/path/to/file.js:line:column`, if the frame represents a call in a user program, or its dependencies.
It is important to note that the string representing the stacktrace is only
generated on **access**: it is lazily generated.
The number of frames captured by the stack trace is bounded by the smaller of
`Error.stackTraceLimit` or the number of available frames on the current event
loop tick.
System-level errors are generated as augmented Error instances, which are detailed
[below](#errors_system_errors).
#### Error.captureStackTrace(targetObject[, constructorOpt])
Creates a `.stack` property on `targetObject`, which when accessed returns
a string representing the location in the program at which `Error.captureStackTrace`
was called.
```javascript
var myObject = {};
Error.captureStackTrace(myObject);
myObject.stack // similar to `new Error().stack`
```
The first line of the trace, instead of being prefixed with `ErrorType:
message`, will be the result of `targetObject.toString()`.
`constructorOpt` optionally accepts a function. If given, all frames above
`constructorOpt`, including `constructorOpt`, will be omitted from the generated
stack trace.
This is useful for hiding implementation details of error generation from the
end user. A common way of using this parameter is to pass the current Error
constructor to it:
```javascript
function MyError() {
Error.captureStackTrace(this, MyError);
}
// without passing MyError to captureStackTrace, the MyError
// frame would should up in the .stack property. by passing
// the constructor, we omit that frame and all frames above it.
new MyError().stack
```
#### Error.stackTraceLimit
Property that determines the number of stack frames collected by a stack trace
(whether generated by `new Error().stack` or `Error.captureStackTrace(obj)`).
The initial value is `10`. It may be set to any valid JavaScript number, which
will effect any stack trace captured *after* the value has been changed. If set
to a non-number value, stack traces will not capture any frames and will report
`undefined` on access.
### Class: RangeError
A subclass of Error that indicates that a provided argument was not within the
set or range of acceptable values for a function; whether that be a numeric
range, or outside the set of options for a given function parameter. An example:
```javascript
require('net').connect(-1); // throws RangeError, port should be > 0 && < 65536
```
io.js will generate and throw RangeError instances *immediately* -- they are a form
of argument validation.
### Class: TypeError
A subclass of Error that indicates that a provided argument is not an allowable
type. For example, passing a function to a parameter which expects a string would
be considered a TypeError.
```javascript
require('url').parse(function() { }); // throws TypeError, since it expected a string
```
io.js will generate and throw TypeError instances *immediately* -- they are a form
of argument validation.
### Class: ReferenceError
A subclass of Error that indicates that an attempt is being made to access a variable
that is not defined. Most commonly it indicates a typo, or an otherwise broken program.
While client code may generate and propagate these errors, in practice only V8 will do
so.
```javascript
doesNotExist; // throws ReferenceError, doesNotExist is not a variable in this program.
```
ReferenceError instances will have an `.arguments` member that is an array containing
one element -- a string representing the variable that was not defined.
```javascript
try {
doesNotExist;
} catch(err) {
err.arguments[0] === 'doesNotExist';
}
```
Unless the userland program is dynamically generating and running code,
ReferenceErrors should always be considered a bug in the program, or its
dependencies.
### Class: SyntaxError
A subclass of Error that indicates that a program is not valid JavaScript.
These errors may only be generated and propagated as a result of code
evaluation. Code evaluation may happen as a result of `eval`, `Function`,
`require`, or [vm](vm.html). These errors are almost always indicative of a broken
program.
```javascript
try {
require("vm").runInThisContext("binary ! isNotOk");
} catch(err) {
// err will be a SyntaxError
}
```
SyntaxErrors are unrecoverable from the context that created them – they may only be caught
by other contexts.
### Exceptions vs. Errors
<!--type=misc-->
A JavaScript "exception" is a value that is thrown as a result of an invalid operation or
as the target of a `throw` statement. While it is not required that these values inherit from
`Error`, all exceptions thrown by io.js or the JavaScript runtime *will* be instances of Error.
Some exceptions are *unrecoverable* at the JavaScript layer. These exceptions will always bring
down the process. These are usually failed `assert()` checks or `abort()` calls in the C++ layer.
## System Errors
System errors are generated in response to a program's runtime environment.
Ideally, they represent operational errors that the program needs to be able to
react to. They are generated at the syscall level: an exhaustive list of error
codes and their meanings is available by running `man 2 intro` on most Unices;
or [online](http://man7.org/linux/man-pages/man2/intro.2.html).
In io.js, system errors are represented as augmented Error objects -- not full
subclasses, but instead an error instance with added members.
### Class: System Error
#### error.syscall
A string representing the [syscall](http://man7.org/linux/man-pages/man2/syscall.2.html) that failed.
#### error.errno
#### error.code
A string representing the error code, which is always `E` followed by capital
letters, and may be referenced in `man 2 intro`.
### Common System Errors
This list is **not exhaustive**, but enumerates many of the common system errors when
writing a io.js program. An exhaustive list may be found [here](http://man7.org/linux/man-pages/man2/intro.2.html).
#### EPERM: Operation not permitted
An attempt was made to perform an operation that requires appropriate
privileges.
#### ENOENT: No such file or directory
Commonly raised by [fs](fs.html) operations; a component of the specified pathname
does not exist -- no entity (file or directory) could be found by the given path.
#### EACCES: Permission denied
An attempt was made to access a file in a way forbidden by its file access
permissions.
#### EEXIST: File exists
An existing file was the target of an operation that required that the target
not exist.
#### ENOTDIR: Not a directory
A component of the given pathname existed, but was not a directory as expected.
Commonly raised by [fs.readdir](fs.html#fs_fs_readdir_path_callback).
#### EISDIR: Is a directory
An operation expected a file, but the given pathname was a directory.
#### EMFILE: Too many open files in system
Maxiumum number of [file descriptors](http://en.wikipedia.org/wiki/File_descriptor) allowable on the system has
been reached, and requests for another descriptor cannot be fulfilled until
at least one has been closed.
Commonly encountered when opening many files at once in parallel, especially
on systems (in particular, OS X) where there is a low file descriptor limit
for processes. To remedy a low limit, run `ulimit -n 2048` in the same shell
that will run the io.js process.
#### EPIPE: Broken pipe
A write on a pipe, socket, or FIFO for which there is no process to read the
data. Commonly encountered at the [net](net.html) and [http](http.html) layers, indicative that
the remote side of the stream being written to has been closed.
#### EADDRINUSE: Address already in use
An attempt to bind a server ([net](net.html), [http](http.html), or [https](https.html)) to a local
address failed due to another server on the local system already occupying
that address.
#### ECONNRESET: Connection reset by peer
A connection was forcibly closed by a peer. This normally results
from a loss of the connection on the remote socket due to a timeout
or reboot. Commonly encountered via the [http](http.html) and [net](net.html) modules.
#### ECONNREFUSED: Connection refused
No connection could be made because the target machine actively refused
it. This usually results from trying to connect to a service that is inactive
on the foreign host.
#### ENOTEMPTY: Directory not empty
A directory with entries was the target of an operation that requires
an empty directory -- usually [fs.unlink](fs.html#fs_fs_unlink_path_callback).
#### ETIMEDOUT: Operation timed out
A connect or send request failed because the connected party did not properly
respond after a period of time. Usually encountered by [http](http.html) or [net](net.html) --
often a sign that a connected socket was not `.end()`'d appropriately.
## Error Propagation and Interception
<!--type=misc-->
All io.js APIs will treat invalid arguments as exceptional -- that is, if passed
invalid arguments, they will *immediately* generate and throw the error as an
exception, even if they are an otherwise asynchronous API.
Synchronous APIs (like
[fs.readFileSync](fs.html#fs_fs_readfilesync_filename_options)) will throw the
error. The act of *throwing* a value (in this case, the error) turns the value
into an **exception**. Exceptions may be caught using the `try { } catch(err)
{ }` construct.
Asynchronous APIs have **two** mechanisms for error propagation; one mechanism
for APIs that represent a single operation, and one for APIs that represent
multiple operations over time.
### Node style callbacks
<!--type=misc-->
Single operation APIs take "node style callbacks" -- a
function provided to the API as an argument. The node style callback takes
at least **one** argument -- `error` -- that will either be `null` (if no error
was encountered) or an `Error` instance. For instance:
```javascript
var fs = require('fs');
fs.readFile('/some/file/that/does-not-exist', function nodeStyleCallback(err, data) {
console.log(err) // Error: ENOENT
console.log(data) // undefined / null
});
fs.readFile('/some/file/that/does-exist', function(err, data) {
console.log(err) // null
console.log(data) // <Buffer: ba dd ca fe>
})
```
Note that `try { } catch(err) { }` **cannot** intercept errors generated by
asynchronous APIs. A common mistake for beginners is to try to use `throw`
inside their node style callback:
```javascript
// THIS WILL NOT WORK:
var fs = require('fs');
try {
fs.readFile('/some/file/that/does-not-exist', function(err, data) {
// mistaken assumption: throwing here...
if (err) {
throw err;
}
});
} catch(err) {
// ... will be caught here -- this is incorrect!
console.log(err); // Error: ENOENT
}
```
This will not work! By the time the node style callback has been called, the
surrounding code (including the `try { } catch(err) { }` will have already
exited. Throwing an error inside a node style callback **will crash the process** in most cases.
If [domains](domain.html) are enabled, they may intercept the thrown error; similarly, if a
handler has been added to `process.on('uncaughtException')`, it will intercept
the error.
### Error events
<!--type=misc-->
The other mechanism for providing errors is the "error" event. This is
typically used by [stream-based](stream.html) and [event emitter-based](events.html#events_class_events_eventemitter) APIs, which
themselves represent a series of asynchronous operations over time (versus a
single operation that may pass or fail). If no "error" event handler is
attached to the source of the error, the error will be thrown. At this point,
it will crash the process as an unhandled exception unless [domains](domain.html) are
employed appropriately or [process.on('uncaughtException')](process.html#process_event_uncaughtexception) has a handler.
```javascript
var net = require('net');
var connection = net.connect('localhost');
// adding an "error" event handler to a stream:
connection.on('error', function(err) {
// if the connection is reset by the server, or if it can't
// connect at all, or on any sort of error encountered by
// the connection, the error will be sent here.
console.error(err);
});
connection.pipe(process.stdout);
```
The "throw when no error handlers are attached behavior" is not limited to APIs
provided by io.js -- even user created event emitters and streams will throw
errors when no error handlers are attached. An example:
```javascript
var events = require('events');
var ee = new events.EventEmitter;
setImmediate(function() {
// this will crash the process because no "error" event
// handler has been added.
ee.emit('error', new Error('This will crash'));
});
```
As with node style callbacks, errors generated this way *cannot* be intercepted
by `try { } catch(err) { }` -- they happen *after* the calling code has already
exited.