Create a block of code, a context or "body" function, which performs some target work, but which has some sort of "set up" and/or "tear down" work to do. In addition to the required main body function, a context can:
- Have a head function that executes immediately prior to the body function
- Have a tail function that executes upon completion of the body function, even if an error occurs
- Have an error that is called when an error happens inside head, body, or tail.
- Manage state which is referenced via the
thiskeyword from within head, body, or tail
Import it as a library:
- Project code is:
MY_Ti1O7GqK78zDPukpS-gnnHlT3Tf0e1 - Script ID is:
1B2L-N3JRuqsXpURb8dyngYZwg0c37mHg7jKtSKYZn5FFjPxH0XM7Ito1
The default identifier is ContextManager, you use the .create function to get the context variable, on which you declare the head, body, and tail as needed.
This package is documented with jsdoc, and so API details are fully documented. You'll notice that the online IDE has limited ability to read jsdoc properly, so please do use the API for a full explanation.
This appscripts library is also an npm module, installable with npm install classroomtechtools/contextmanager and imported via import {ContextManager} from '@classroomtechtools/contextmanager'.
For completeness, this library is tested via unit testing on a node environment. See bottom for output.
/**
* A simple (and useless) context manager that illustrates patterns
*/
function myFunction () {
const context = ContextManager.create();
context.head = function (param) {
// this will be state, by default just an object
this.inHead = true;
};
context.body = function (param) {
this.inBody = true;
return param;
};
context.tail = function (param) {
this.inTail = true;
}
const result = context.execute("echo");
Logger.log(result);
// echo
Logger.log(context.state);
// {inHead: true, inTail: true, inBody: true};
}/**
* A simple (and vastly more useful) context manager
* that puts the body function inside of a script-level lock
* @see https://classroomtechtools.github.io/ContextManager/global.html#usingWaitLock
*/
function myFunction () {
const context = ContextManager.usingWaitLock(500, 'script');
context.body = function (param) {
return param + 1;
};
const result = context.execute(1);
Logger.log(result);
// 2
}In many cases, you may not need to be aware of the below minutiae, but in provided just in case.
By default, this is the order in which execution occurs. Please see below example for more clarity:
1is head function2is body function3is tail function-1is error function
Regular sequence (with no error occurring) 1 2 3 (head body tail)
If error happens in head function, sequence is 1 -1 3 (head error tail, but not body).
If an error occurs in body function, sequence is:
1 2 -1 3 (head body error tail)
If an error occurs in the tail function, sequence is:
1 2 3, -1 (head, body, tail, error)
Note: If an error occurs in the error function, then the sequence will be interrupted at the -1 stage with no further execution; that is the intended behaviour.
Error Handling
If the error function returns null, the error is "muted", which means it is not raised as an error, same behaviour as muteHttpExceptions when using UrlFetchApp.fetch. Return null in the error function means that any error, if it occurs, it is returned by .execute. If the error function has not been declared, the default behaviour is that it is not muted, i.e. it is raised.
Body returning value
The body can return some value which is returned from .execute. If an error occurs in body function and the error function returns null, the error itself is returned instead.
If an error occurs in the tail function (and thus the body function has successfully return some value or undefined if no explicit return) and the error function returns null, then the returned error object is given additional property ctx.body.result.
A common pattern where this library would be useful is when you are working with spreadsheets and the LockService in tandem. In order to prevent your own script from overwriting your own updates, you need to create a lock at the script level, then do your writes, then you need to call SpreadsheetApp.flush() to write changes. When you enable the lock, you might get an error (as waitLock throws an error if it was unable to establish a lock).
/*
/* Update spreadsheet
/* @returns error message if error happens, else null
*/
function myFunction () {
const context = ContextManager.create(); // creates instance of class
// define head function
context.head = function () {
console.log(1);
this.error = null;
this.lock = LockService.getScriptLock();
this.lock.waitLock(300); // timeout of 300
};
// now create body function with first parameter
context.body = function () {
console.log(2);
SpreadsheetApp... // update the spreadsheet, inside a lock
this.updated = true;
return this.error;
};
// define tail function
context.tail = function () {
console.log(3);
SpreadsheetApp.flush();
this.lock.releaseLock();
};
// define what happens when an error is thrown
context.error = function (err) {
console.log(-1);
this.error = err;
return null; // "instructs context manager to 'mute' the error"
};
return context.execute();
// ^--- this actually executes it all
}The following silly example illustrates how this holds state throughout each of the functions, and how you can use .execute to pass parameters which can change how the body function works:
function myFunction () {
// initialize
let ctx = ContextManager.create();
// save an array to state
ctx.head = function () {
this.noun = "World";
};
// execute the main body, head and tail will be called
ctx.body = function (delim) {
this.expression = `Hello${delim}${this.noun}`;
};
// output the array to the logger
ctx.tail = function () {
Logger.log(this.expression);
};
ctx.execute(', ');
// Logger outputs "Hello, World"
ctx.execute(' ');
// Logger outputs "Hello World" (with no comma)
}By default, this is just a regular object. If you want it to be something else, then you can set its state ahead of time:
function myFunction () {
let ctx = ContextManager.create({state: []}); // pass array which is state object
ctx.head = function () {
this.push('heading');
};
ctx.tail = function () {
this.push('tailing');
Logger.log(this.log.join('\n'));
};
ctx.error = function () {
this.push('See no error, hear no error');
return null; // return null mutes the error
};
ctx.param = "World";
ctx.with(function (text) {
this.push('Inside body');
throw Error("Error here, but does not actually error out");
});
}Output:
heading
Inside body
See no error, hear no error
tailing
Context managers are a concept in Python that is really quite useful. In my case, I use them to implement unit testing, as a unit test may possibly fail, but we want the code to continue executing.
✔ sequence › Sequence for when error occurs in head is 1, -1, 3
✔ sequence › Sequence for when error occurs in body is 1, 2, -1, 3
✔ sequence › Sequence for when error occurs in tail is 1, 2, 3, -1
✔ declarative › set context.body and context.execute for multiple invocations
✔ declarative › this keyword holds state, is available via state property
✔ declarative › pass array as state to contructor
✔ declarative › state can be set, if null reverts to the default object
✔ declarative › subclassing to get different defaultObject
✔ declarative › Saves state between head and tail calls, return this
✔ declarative › param property is passed to with method as parameter
✔ declarative › withLock
✔ declarative › param is sent to head, body, and tail
✔ declarative › create() with callbacks
✔ errorHandling › default error handler does not mute error
✔ errorHandling › default error handler does not mute error in head
✔ errorHandling › error handler mutes error by returning null
✔ errorHandling › if error occurs in head, body, or tail and is muted, the error obj is returned
✔ errorHandling › if error occurs in tail but body returned something, it can be found in returned error as result property
✔ failures › usingWaitLock throws error on incorrect guard value
✔ failures › usingWaitLock throws error on incorrect parameter