I'm learning/experimenting with Flink, and I'm observing some unexpected behavior with the DataStream join, and would like to understand what is happening...
Let's say I have two streams with 10 records each, which I want to join on a id field. Let's assume that for each record in one stream had a matching one in the other, and the IDs are unique in each stream. Let's also say I have to use a global window (requirement).
Join using DataStream API (my simplified code in Scala):
val stream1 = ... // from a Kafka topic on my local machine (I tried with and without .keyBy)
val stream2 = ...
stream1
.join(stream2)
.where(_.id).equalTo(_.id)
.window(GlobalWindows.create()) // assume this is a requirement
.trigger(CountTrigger.of(1))
.apply {
(row1, row2) => // ...
}
.print()
Result:
Everything is printed as expected, each record from the first stream joined with a record from the second one.
However:
If I re-send one of the records (say, with an updated field) from one of the stream to that stream, two duplicate join events get emitted 😞
If I repeat that operation (with or without updated field), I will get 3 emitted events, then 4, 5, etc... 😞
Could someone in the Flink community explain why this is happening? I would have expected only 1 event emitted each time. Is it possible to achieve this with a global window?
In comparison, the Flink Table API behaves as expected in that same scenario, but for my project I'm more interested in the DataStream API.
Example with Table API, which worked as expected:
tableEnv
.sqlQuery(
"""
|SELECT *
| FROM stream1
| JOIN stream2
| ON stream1.id = stream2.id
""".stripMargin)
.toRetractStream[Row]
.filter(_._1) // just keep the inserts
.map(...)
.print() // works as expected, after re-sending updated records
Thank you,
Nicolas
The issue is that records are never removed from your global window. So you trigger the join operation on the global window, whenever a new record has arrived, but the old records are still present.
Thus, to get it running in your case, you'd need to implement a custom evictor. I expanded your example in a minimal working example and added the evictor, which I will explain after the snippet.
val data1 = List(
(1L, "myId-1"),
(2L, "myId-2"),
(5L, "myId-1"),
(9L, "myId-1"))
val data2 = List(
(3L, "myId-1", "myValue-A"))
val stream1 = env.fromCollection(data1)
val stream2 = env.fromCollection(data2)
stream1.join(stream2)
.where(_._2).equalTo(_._2)
.window(GlobalWindows.create()) // assume this is a requirement
.trigger(CountTrigger.of(1))
.evictor(new Evictor[CoGroupedStreams.TaggedUnion[(Long, String), (Long, String, String)], GlobalWindow](){
override def evictBefore(elements: lang.Iterable[TimestampedValue[CoGroupedStreams.TaggedUnion[(Long, String), (Long, String, String)]]], size: Int, window: GlobalWindow, evictorContext: Evictor.EvictorContext): Unit = {}
override def evictAfter(elements: lang.Iterable[TimestampedValue[CoGroupedStreams.TaggedUnion[(Long, String), (Long, String, String)]]], size: Int, window: GlobalWindow, evictorContext: Evictor.EvictorContext): Unit = {
import scala.collection.JavaConverters._
val lastInputTwoIndex = elements.asScala.zipWithIndex.filter(e => e._1.getValue.isTwo).lastOption.map(_._2).getOrElse(-1)
if (lastInputTwoIndex == -1) {
println("Waiting for the lookup value before evicting")
return
}
val iterator = elements.iterator()
for (index <- 0 until size) {
val cur = iterator.next()
if (index != lastInputTwoIndex) {
println(s"evicting ${cur.getValue.getOne}/${cur.getValue.getTwo}")
iterator.remove()
}
}
}
})
.apply((r, l) => (r, l))
.print()
The evictor will be applied after the window function (join in this case) has been applied. It's not entirely clear how your use case exactly should work in case you have multiple entries in the second input, but for now, the evictor only works with single entries.
Whenever a new element comes into the window, the window function is immediately triggered (count = 1). Then the join is evaluated with all elements having the same key. Afterwards, to avoid duplicate outputs, we remove all entries from the first input in the current window. Since, the second input may arrive after the first inputs, no eviction is performed, when the second input is empty. Note that my scala is quite rusty; you will be able to write it in a much nicer way. The output of a run is:
Waiting for the lookup value before evicting
Waiting for the lookup value before evicting
Waiting for the lookup value before evicting
Waiting for the lookup value before evicting
4> ((1,myId-1),(3,myId-1,myValue-A))
4> ((5,myId-1),(3,myId-1,myValue-A))
4> ((9,myId-1),(3,myId-1,myValue-A))
evicting (1,myId-1)/null
evicting (5,myId-1)/null
evicting (9,myId-1)/null
A final remark: if the table API offers already a concise way of doing what you want, I'd stick to it and then convert it to a DataStream when needed.
I want to find pattern of events that follow
Inner pattern is:
Have the same value for key "sensorArea".
Have different value for key "customerId".
Are within 5 seconds from each other.
And this pattern needs to
Emit "alert" only if previous happens 3 or more times.
I wrote something but I know for sure it is not complete.
Two Questions
I need to access the previous event fields when I'm in the "next" pattern, how can I do that without using the ctx command because it is heavy..
My code brings weird result - this is my input
and my output is
3> {first=[Customer[timestamp=50,customerId=111,toAdd=2,sensorData=33]], second=[Customer[timestamp=100,customerId=222,toAdd=2,sensorData=33], Customer[timestamp=600,customerId=333,toAdd=2,sensorData=33]]}
even though my desired output should be all first six events (users 111/222 and sensor are 33 and then 44 and then 55
Pattern<Customer, ?> sameUserDifferentSensor = Pattern.<Customer>begin("first", skipStrategy)
.followedBy("second").where(new IterativeCondition<Customer>() {
#Override
public boolean filter(Customer currCustomerEvent, Context<Customer> ctx) throws Exception {
List<Customer> firstPatternEvents = Lists.newArrayList(ctx.getEventsForPattern("first"));
int i = firstPatternEvents.size();
int currSensorData = currCustomerEvent.getSensorData();
int prevSensorData = firstPatternEvents.get(i-1).getSensorData();
int currCustomerId = currCustomerEvent.getCustomerId();
int prevCustomerId = firstPatternEvents.get(i-1).getCustomerId();
return currSensorData==prevSensorData && currCustomerId!=prevCustomerId;
}
})
.within(Time.seconds(5))
.timesOrMore(3);
PatternStream<Customer> sameUserDifferentSensorPatternStream = CEP.pattern(customerStream, sameUserDifferentSensor);
DataStream<String> alerts1 = sameUserDifferentSensorPatternStream.select((PatternSelectFunction<Customer, String>) Object::toString);
You will have an easier time if you first key the stream by the sensorArea. They you will be pattern matching on streams where all of the events are for a single sensorArea, which will make the pattern easier to express, and the matching more efficient.
You can't avoid using an iterative condition and the ctx, but it should be less expensive after keying the stream.
Also, your code example doesn't match the text description. The text says "within 5 seconds" and "3 or more times", while the code has within(Time.seconds(2)) and timesOrMore(2).
My Google Apps Script is iterating through the user's Google Drive files and copying and sometimes moving files to other folders. The script is always stopped after certain minutes with no error message in the log.
EDITOR's NOTE: The time limit have varied over the time and might vary between "consumer" (free) and "Workspace" (paid) accounts but as of December 2022 most of the answers are still valid.
I am sorting tens or sometimes thousands files in one run.
Are there any settings or workarounds?
One thing you could do (this of course depends on what you are trying to accomplish) is:
Store the necessary information (i.e. like a loop counter) in a spreadsheet or another permanent store(i.e. ScriptProperties).
Have your script terminate every five minutes or so.
Set up a time driven trigger to run the script every five minutes(or create a trigger programmatically using the Script service).
On each run read the saved data from the permanent store you've used and continue to run the script from where it left off.
This is not a one-size-fit-all solution, if you post your code people would be able to better assist you.
Here is a simplified code excerpt from a script that I use every day:
function runMe() {
var startTime= (new Date()).getTime();
//do some work here
var scriptProperties = PropertiesService.getScriptProperties();
var startRow= scriptProperties.getProperty('start_row');
for(var ii = startRow; ii <= size; ii++) {
var currTime = (new Date()).getTime();
if(currTime - startTime >= MAX_RUNNING_TIME) {
scriptProperties.setProperty("start_row", ii);
ScriptApp.newTrigger("runMe")
.timeBased()
.at(new Date(currTime+REASONABLE_TIME_TO_WAIT))
.create();
break;
} else {
doSomeWork();
}
}
//do some more work here
}
NOTE#1: The variable REASONABLE_TIME_TO_WAIT should be large enough for the new trigger to fire. (I set it to 5 minutes but I think it could be less than that).
NOTE#2: doSomeWork() must be a function that executes relatively quick( I would say less than 1 minute ).
NOTE#3 : Google has deprecated Script Properties, and introduced Properties Service in its stead. The function has been modified accordingly.
NOTE#4: 2nd time when the function is called, it takes the ith value of for loop as a string. so you have to convert it into an integer
Quotas
The maximum execution time for a single script is 6 mins / execution
- https://developers.google.com/apps-script/guides/services/quotas
But there are other limitations to familiarize yourself with. For example, you're only allowed a total trigger runtime of 1 hour / day, so you can't just break up a long function into 12 different 5 minute blocks.
Optimization
That said, there are very few reasons why you'd really need to take six minutes to execute. JavaScript should have no problem sorting thousands of rows of data in a couple seconds. What's likely hurting your performance are service calls to Google Apps itself.
You can write scripts to take maximum advantage of the built-in caching, by minimizing the number of reads and writes. Alternating read and write commands is slow. To speed up a script, read all data into an array with one command, perform any operations on the data in the array, and write the data out with one command.
- https://developers.google.com/apps-script/best_practices
Batching
The best thing you can possibly do is reduce the number of service calls. Google enables this by allowing batch versions of most of their API calls.
As a trivial example, Instead of this:
for (var i = 1; i <= 100; i++) {
SpreadsheetApp.getActiveSheet().deleteRow(i);
}
Do this:
SpreadsheetApp.getActiveSheet().deleteRows(i, 100);
In the first loop, not only did you need 100 calls to deleteRow on the sheet, but you also needed to get the active sheet 100 times as well. The second variation should perform several orders of magnitude better than the first.
Interweaving Reads and Writes
Additionally, you should also be very careful to not go back and forth frequently between reading and writing. Not only will you lose potential gains in batch operations, but Google won't be able to use its built-in caching.
Every time you do a read, we must first empty (commit) the write cache to ensure that you're reading the latest data (you can force a write of the cache by calling SpreadsheetApp.flush()). Likewise, every time you do a write, we have to throw away the read cache because it's no longer valid. Therefore if you can avoid interleaving reads and writes, you'll get full benefit of the cache.
- http://googleappsscript.blogspot.com/2010/06/optimizing-spreadsheet-operations.html
For example, instead of this:
sheet.getRange("A1").setValue(1);
sheet.getRange("B1").setValue(2);
sheet.getRange("C1").setValue(3);
sheet.getRange("D1").setValue(4);
Do this:
sheet.getRange("A1:D1").setValues([[1,2,3,4]]);
Chaining Function Calls
As a last resort, if your function really can't finish in under six minutes, you can chain together calls or break up your function to work on a smaller segment of data.
You can store data in the Cache Service (temporary) or Properties Service (permanent) buckets for retrieval across executions (since Google Apps Scripts has a stateless execution).
If you want to kick off another event, you can create your own trigger with the Trigger Builder Class or setup a recurring trigger on a tight time table.
Also, try to minimize the amount of calls to google services. For example, if you want to change a range of cells in the spreadsheets, don't read each one, mutate it and store it back.
Instead read the whole range (using Range.getValues()) into memory, mutate it and store all of it at once (using Range.setValues()).
This should save you a lot of execution time.
Anton Soradoi's answer seems OK but consider using Cache Service instead of storing data into a temporary sheet.
function getRssFeed() {
var cache = CacheService.getPublicCache();
var cached = cache.get("rss-feed-contents");
if (cached != null) {
return cached;
}
var result = UrlFetchApp.fetch("http://example.com/my-slow-rss-feed.xml"); // takes 20 seconds
var contents = result.getContentText();
cache.put("rss-feed-contents", contents, 1500); // cache for 25 minutes
return contents;
}
Also note that as of April 2014 the limitation of script runtime is 6 minutes.
G Suite Business / Enterprise / Education and Early Access users:
As of August 2018, max script runtime is now set to 30 minutes for these users.
Figure out a way to split up your work so it takes less than 6 minutes, as that's the limit for any script. On the first pass, you can iterate and store the list of files and folders in a spreadsheet and add a time-driven trigger for part 2.
In part 2, delete each entry in the list as you process it. When there are no items in the list, delete the trigger.
This is how I'm processing a sheet of about 1500 rows that gets spread to about a dozen different spreadsheets. Because of the number of calls to spreadsheets, it times out, but continues when the trigger runs again.
I have used the ScriptDB to save my place while processing a large amount of information in a loop. The script can/does exceed the 5 minute limit. By updating the ScriptDb during each run, the script can read the state from the db and pick up where it left off until all processing is complete. Give this strategy a try and I think you'll be pleased with the results.
If you are using G Suite Business or Enterprise edition.
You can register early access for App Maker after App maker enabled your script run runtime will increase run time from 6 minutes to 30 minutes :)
More details about app maker Click here
Here's an approach based very heavily on Dmitry Kostyuk's absolutely excellent article on the subject.
It differs in that it doesn't attempt to time execution and exit gracefully. Rather, it deliberately spawns a new thread every minute, and lets them run until they are timed out by Google. This gets round the maximum execution time limit, and speeds things up by running processing in several threads in parallel. (This speeds things up even if you are not hitting execution time limits.)
It tracks the task status in script properties, plus a semaphore to ensure no two threads are editing the task status at any one time. (It uses several properties as they are limited to 9k each.)
I have tried to mimick the Google Apps Script iterator.next() API, but cannot use iterator.hasNext() as that would not be thread-safe (see TOCTOU). It uses a couple of facade classes at the bottom.
I would be immensely grateful for any suggestions. This is working well for me, halving the processing time by spawning three parallel threads to run through a directory of documents. You could spawn 20 within quota, but this was ample for my use case.
The class is designed to be drop-in, usable for any purpose without modification. The only thing the user must do is when processing a file, delete any outputs from prior, timed out attempts. The iterator will return a given fileId more than once if a processing task is timed out by Google before it completes.
To silence the logging, it all goes through the log() function at the bottom.
This is how you use it:
const main = () => {
const srcFolder = DriveApp.getFoldersByName('source folder',).next()
const processingMessage = processDocuments(srcFolder, 'spawnConverter')
log('main() finished with message', processingMessage)
}
const spawnConverter = e => {
const processingMessage = processDocuments()
log('spawnConverter() finished with message', processingMessage)
}
const processDocuments = (folder = null, spawnFunction = null) => {
// folder and spawnFunction are only passed the first time we trigger this function,
// threads spawned by triggers pass nothing.
// 10,000 is the maximum number of milliseconds a file can take to process.
const pfi = new ParallelFileIterator(10000, MimeType.GOOGLE_DOCS, folder, spawnFunction)
let fileId = pfi.nextId()
const doneDocs = []
while (fileId) {
const fileRelativePath = pfi.getFileRelativePath(fileId)
const doc = DocumentApp.openById(fileId)
const mc = MarkupConverter(doc)
// This is my time-consuming task:
const mdContent = mc.asMarkdown(doc)
pfi.completed(fileId)
doneDocs.push([...fileRelativePath, doc.getName() + '.md'].join('/'))
fileId = pfi.nextId()
}
return ('This thread did:\r' + doneDocs.join('\r'))
}
Here's the code:
const ParallelFileIterator = (function() {
/**
* Scans a folder, depth first, and returns a file at a time of the given mimeType.
* Uses ScriptProperties so that this class can be used to process files by many threads in parallel.
* It is the responsibility of the caller to tidy up artifacts left behind by processing threads that were timed out before completion.
* This class will repeatedly dispatch a file until .completed(fileId) is called.
* It will wait maxDurationOneFileMs before re-dispatching a file.
* Note that Google Apps kills scripts after 6 mins, or 30 mins if you're using a Workspace account, or 45 seconds for a simple trigger, and permits max 30
* scripts in parallel, 20 triggers per script, and 90 mins or 6hrs of total trigger runtime depending if you're using a Workspace account.
* Ref: https://developers.google.com/apps-script/guides/services/quotas
maxDurationOneFileMs, mimeType, parentFolder=null, spawnFunction=null
* #param {Number} maxDurationOneFileMs A generous estimate of the longest a file can take to process.
* #param {string} mimeType The mimeType of the files required.
* #param {Folder} parentFolder The top folder containing all the files to process. Only passed in by the first thread. Later spawned threads pass null (the files have already been listed and stored in properties).
* #param {string} spawnFunction The name of the function that will spawn new processing threads. Only passed in by the first thread. Later spawned threads pass null (a trigger can't create a trigger).
*/
class ParallelFileIterator {
constructor(
maxDurationOneFileMs,
mimeType,
parentFolder = null,
spawnFunction = null,
) {
log(
'Enter ParallelFileIterator constructor',
maxDurationOneFileMs,
mimeType,
spawnFunction,
parentFolder ? parentFolder.getName() : null,
)
// singleton
if (ParallelFileIterator.instance) return ParallelFileIterator.instance
if (parentFolder) {
_cleanUp()
const t0 = Now.asTimestamp()
_getPropsLock(maxDurationOneFileMs)
const t1 = Now.asTimestamp()
const { fileIds, fileRelativePaths } = _catalogFiles(
parentFolder,
mimeType,
)
const t2 = Now.asTimestamp()
_setQueues(fileIds, [])
const t3 = Now.asTimestamp()
this.fileRelativePaths = fileRelativePaths
ScriptProps.setAsJson(_propsKeyFileRelativePaths, fileRelativePaths)
const t4 = Now.asTimestamp()
_releasePropsLock()
const t5 = Now.asTimestamp()
if (spawnFunction) {
// only triggered on the first thread
const trigger = Trigger.create(spawnFunction, 1)
log(
`Trigger once per minute: UniqueId: ${trigger.getUniqueId()}, EventType: ${trigger.getEventType()}, HandlerFunction: ${trigger.getHandlerFunction()}, TriggerSource: ${trigger.getTriggerSource()}, TriggerSourceId: ${trigger.getTriggerSourceId()}.`,
)
}
log(
`PFI instantiated for the first time, has found ${
fileIds.length
} documents to process. getPropsLock took ${t1 -
t0}ms, _catalogFiles took ${t2 - t1}ms, setQueues took ${t3 -
t2}ms, setAsJson took ${t4 - t3}ms, releasePropsLock took ${t5 -
t4}ms, trigger creation took ${Now.asTimestamp() - t5}ms.`,
)
} else {
const t0 = Now.asTimestamp()
// wait for first thread to set up Properties
while (!ScriptProps.getJson(_propsKeyFileRelativePaths)) {
Utilities.sleep(250)
}
this.fileRelativePaths = ScriptProps.getJson(_propsKeyFileRelativePaths)
const t1 = Now.asTimestamp()
log(
`PFI instantiated again to run in parallel. getJson(paths) took ${t1 -
t0}ms`,
)
spawnFunction
}
_internals.set(this, { maxDurationOneFileMs: maxDurationOneFileMs })
// to get: _internal(this, 'maxDurationOneFileMs')
ParallelFileIterator.instance = this
return ParallelFileIterator.instance
}
nextId() {
// returns false if there are no more documents
const maxDurationOneFileMs = _internals.get(this).maxDurationOneFileMs
_getPropsLock(maxDurationOneFileMs)
let { pending, dispatched } = _getQueues()
log(
`PFI.nextId: ${pending.length} files pending, ${
dispatched.length
} dispatched, ${Object.keys(this.fileRelativePaths).length -
pending.length -
dispatched.length} completed.`,
)
if (pending.length) {
// get first pending Id, (ie, deepest first)
const nextId = pending.shift()
dispatched.push([nextId, Now.asTimestamp()])
_setQueues(pending, dispatched)
_releasePropsLock()
return nextId
} else if (dispatched.length) {
log(`PFI.nextId: Get first dispatched Id, (ie, oldest first)`)
let startTime = dispatched[0][1]
let timeToTimeout = startTime + maxDurationOneFileMs - Now.asTimestamp()
while (dispatched.length && timeToTimeout > 0) {
log(
`PFI.nextId: None are pending, and the oldest dispatched one hasn't yet timed out, so wait ${timeToTimeout}ms to see if it will`,
)
_releasePropsLock()
Utilities.sleep(timeToTimeout + 500)
_getPropsLock(maxDurationOneFileMs)
;({ pending, dispatched } = _getQueues())
if (pending && dispatched) {
if (dispatched.length) {
startTime = dispatched[0][1]
timeToTimeout =
startTime + maxDurationOneFileMs - Now.asTimestamp()
}
}
}
// We currently still have the PropsLock
if (dispatched.length) {
const nextId = dispatched.shift()[0]
log(
`PFI.nextId: Document id ${nextId} has timed out; reset start time, move to back of queue, and re-dispatch`,
)
dispatched.push([nextId, Now.asTimestamp()])
_setQueues(pending, dispatched)
_releasePropsLock()
return nextId
}
}
log(`PFI.nextId: Both queues empty, all done!`)
;({ pending, dispatched } = _getQueues())
if (pending.length || dispatched.length) {
log(
"ERROR: All documents should be completed, but they're not. Giving up.",
pending,
dispatched,
)
}
_cleanUp()
return false
}
completed(fileId) {
_getPropsLock(_internals.get(this).maxDurationOneFileMs)
const { pending, dispatched } = _getQueues()
const newDispatched = dispatched.filter(el => el[0] !== fileId)
if (dispatched.length !== newDispatched.length + 1) {
log(
'ERROR: A document was completed, but not found in the dispatched list.',
fileId,
pending,
dispatched,
)
}
if (pending.length || newDispatched.length) {
_setQueues(pending, newDispatched)
_releasePropsLock()
} else {
log(`PFI.completed: Both queues empty, all done!`)
_cleanUp()
}
}
getFileRelativePath(fileId) {
return this.fileRelativePaths[fileId]
}
}
// ============= PRIVATE MEMBERS ============= //
const _propsKeyLock = 'PropertiesLock'
const _propsKeyDispatched = 'Dispatched'
const _propsKeyPending = 'Pending'
const _propsKeyFileRelativePaths = 'FileRelativePaths'
// Not really necessary for a singleton, but in case code is changed later
var _internals = new WeakMap()
const _cleanUp = (exceptProp = null) => {
log('Enter _cleanUp', exceptProp)
Trigger.deleteAll()
if (exceptProp) {
ScriptProps.deleteAllExcept(exceptProp)
} else {
ScriptProps.deleteAll()
}
}
const _catalogFiles = (folder, mimeType, relativePath = []) => {
// returns IDs of all matching files in folder, depth first
log(
'Enter _catalogFiles',
folder.getName(),
mimeType,
relativePath.join('/'),
)
let fileIds = []
let fileRelativePaths = {}
const folders = folder.getFolders()
let subFolder
while (folders.hasNext()) {
subFolder = folders.next()
const results = _catalogFiles(subFolder, mimeType, [
...relativePath,
subFolder.getName(),
])
fileIds = fileIds.concat(results.fileIds)
fileRelativePaths = { ...fileRelativePaths, ...results.fileRelativePaths }
}
const files = folder.getFilesByType(mimeType)
while (files.hasNext()) {
const fileId = files.next().getId()
fileIds.push(fileId)
fileRelativePaths[fileId] = relativePath
}
return { fileIds: fileIds, fileRelativePaths: fileRelativePaths }
}
const _getQueues = () => {
const pending = ScriptProps.getJson(_propsKeyPending)
const dispatched = ScriptProps.getJson(_propsKeyDispatched)
log('Exit _getQueues', pending, dispatched)
// Note: Empty lists in Javascript are truthy, but if Properties have been deleted by another thread they'll be null here, which are falsey
return { pending: pending || [], dispatched: dispatched || [] }
}
const _setQueues = (pending, dispatched) => {
log('Enter _setQueues', pending, dispatched)
ScriptProps.setAsJson(_propsKeyPending, pending)
ScriptProps.setAsJson(_propsKeyDispatched, dispatched)
}
const _getPropsLock = maxDurationOneFileMs => {
// will block until lock available or lock times out (because a script may be killed while holding a lock)
const t0 = Now.asTimestamp()
while (
ScriptProps.getNum(_propsKeyLock) + maxDurationOneFileMs >
Now.asTimestamp()
) {
Utilities.sleep(2000)
}
ScriptProps.set(_propsKeyLock, Now.asTimestamp())
log(`Exit _getPropsLock: took ${Now.asTimestamp() - t0}ms`)
}
const _releasePropsLock = () => {
ScriptProps.delete(_propsKeyLock)
log('Exit _releasePropsLock')
}
return ParallelFileIterator
})()
const log = (...args) => {
// easier to turn off, json harder to read but easier to hack with
console.log(args.map(arg => JSON.stringify(arg)).join(';'))
}
class Trigger {
// Script triggering facade
static create(functionName, everyMinutes) {
return ScriptApp.newTrigger(functionName)
.timeBased()
.everyMinutes(everyMinutes)
.create()
}
static delete(e) {
if (typeof e !== 'object') return log(`${e} is not an event object`)
if (!e.triggerUid)
return log(`${JSON.stringify(e)} doesn't have a triggerUid`)
ScriptApp.getProjectTriggers().forEach(trigger => {
if (trigger.getUniqueId() === e.triggerUid) {
log('deleting trigger', e.triggerUid)
return ScriptApp.delete(trigger)
}
})
}
static deleteAll() {
// Deletes all triggers in the current project.
var triggers = ScriptApp.getProjectTriggers()
for (var i = 0; i < triggers.length; i++) {
ScriptApp.deleteTrigger(triggers[i])
}
}
}
class ScriptProps {
// properties facade
static set(key, value) {
if (value === null || value === undefined) {
ScriptProps.delete(key)
} else {
PropertiesService.getScriptProperties().setProperty(key, value)
}
}
static getStr(key) {
return PropertiesService.getScriptProperties().getProperty(key)
}
static getNum(key) {
// missing key returns Number(null), ie, 0
return Number(ScriptProps.getStr(key))
}
static setAsJson(key, value) {
return ScriptProps.set(key, JSON.stringify(value))
}
static getJson(key) {
return JSON.parse(ScriptProps.getStr(key))
}
static delete(key) {
PropertiesService.getScriptProperties().deleteProperty(key)
}
static deleteAll() {
PropertiesService.getScriptProperties().deleteAllProperties()
}
static deleteAllExcept(key) {
PropertiesService.getScriptProperties()
.getKeys()
.forEach(curKey => {
if (curKey !== key) ScriptProps.delete(key)
})
}
}
If you're a business customer, you can now sign up for Early Access to App Maker, which includes Flexible Quotas.
Under the flexible quota system, such hard quota limits are removed. Scripts do not stop when they reach a quota limit. Rather, they are delayed until quota becomes available, at which point the script execution resumes. Once quotas begin being used, they are refilled at a regular rate. For reasonable usage, script delays are rare.
If you are using G Suite as a Business, Enterprise or EDU customer the execution time for running scripts is set to:
30 min / execution
See: https://developers.google.com/apps-script/guides/services/quotas
The idea would be to exit gracefully from the script, save your progress, create a trigger to start again from where you left off, repeat as many times as necessary and then once finished clean up the trigger and any temporary files.
Here is a detailed article on this very topic.
As many people mentioned, the generic solution to this problem is to execute your method across multiple sessions. I found it to be a common problem that I have a bunch of iterations I need to loop over, and I don't want the hassle of writing/maintaining the boilerplate of creating new sessions.
Therefore I created a general solution:
/**
* Executes the given function across multiple sessions to ensure there are no timeouts.
*
* See https://stackoverflow.com/a/71089403.
*
* #param {Int} items - The items to iterate over.
* #param {function(Int)} fn - The function to execute each time. Takes in an item from `items`.
* #param {String} resumeFunctionName - The name of the function (without arguments) to run between sessions. Typically this is the same name of the function that called this method.
* #param {Int} maxRunningTimeInSecs - The maximum number of seconds a script should be able to run. After this amount, it will start a new session. Note: This must be set to less than the actual timeout as defined in https://developers.google.com/apps-script/guides/services/quotas (e.g. 6 minutes), otherwise it can't set up the next call.
* #param {Int} timeBetweenIterationsInSeconds - The amount of time between iterations of sessions. Note that Google Apps Script won't honor this 100%, as if you choose a 1 second delay, it may actually take a minute or two before it actually executes.
*/
function iterateAcrossSessions(items, fn, resumeFunctionName, maxRunningTimeInSeconds = 5 * 60, timeBetweenIterationsInSeconds = 1) {
const PROPERTY_NAME = 'iterateAcrossSessions_index';
let scriptProperties = PropertiesService.getScriptProperties();
let startTime = (new Date()).getTime();
let startIndex = parseInt(scriptProperties.getProperty(PROPERTY_NAME));
if (Number.isNaN(startIndex)) {
startIndex = 0;
}
for (let i = startIndex; i < items.length; i++) {
console.info(`[iterateAcrossSessions] Executing for i = ${i}.`)
fn(items[i]);
let currentTime = (new Date()).getTime();
let elapsedTime = currentTime - startTime;
let maxRunningTimeInMilliseconds = maxRunningTimeInSeconds * 1000;
if (maxRunningTimeInMilliseconds <= elapsedTime) {
let newTime = new Date(currentTime + timeBetweenIterationsInSeconds * 1000);
console.info(`[iterateAcrossSessions] Creating new session for i = ${i+1} at ${newTime}, since elapsed time was ${elapsedTime}.`);
scriptProperties.setProperty(PROPERTY_NAME, i+1);
ScriptApp.newTrigger(resumeFunctionName).timeBased().at(newTime).create();
return;
}
}
console.log(`[iterateAcrossSessions] Done iterating over items.`);
// Reset the property here to ensure that the execution loop could be restarted.
scriptProperties.deleteProperty(PROPERTY_NAME);
}
You can now use this pretty easily like so:
let ITEMS = ['A', 'B', 'C'];
function execute() {
iterateAcrossSessions(
ITEMS,
(item) => {
console.log(`Hello world ${item}`);
},
"execute");
}
It'll automatically execute the internal lambda for each value in ITEMS, seamlessly spreading across sessions as needed.
For example, if you use a 0-second maxRunningTime it would run across 4 sessions with the following outputs:
[iterateAcrossSessions] Executing for i = 0.
Hello world A
[iterateAcrossSessions] Creating new session for i = 1.
[iterateAcrossSessions] Executing for i = 1.
Hello world B
[iterateAcrossSessions] Creating new session for i = 2.
[iterateAcrossSessions] Executing for i = 2.
Hello world C
[iterateAcrossSessions] Creating new session for i = 3.
[iterateAcrossSessions] Done iterating over items.