A deno runtime for AWS Lambda.
Deploy deno code via SAR application (see quick start), SAM, serverless, or bundle it yourself.
Define a handler function, for example:
// hello.ts
import {
APIGatewayProxyEventV2,
APIGatewayProxyResultV2,
Context,
} from "https://deno.land/x/lambda/mod.ts";
export async function handler(
event: APIGatewayProxyEventV2,
context: Context,
): Promise<APIGatewayProxyResultV2> {
return {
body: `Welcome to deno ${Deno.version.deno} ๐ฆ`,
headers: { "content-type": "text/html;charset=utf8" },
statusCode: 200,
};
}
Here the handler is hello.handler
but this can be configured by the Handler
setting.
The following environment variables can be set to change deno-lambda's behavior:
HANDLER_EXT
to set supported extension of handler file e.g.js
orbundle.js
(defaultts
).DENO_CONFIG
so deno runs with--config=$DENO_CONFIG
.DENO_DIR
so deno runs withDENO_DIR=$DENO_DIR deno ...
.DENO_IMPORTMAP
so deno runs with--importmap=$DENO_IMPORTMAP
.DENO_LOCATION
so deno runs with--location=$DENO_LOCATION
.DENO_LOCK
so deno runs with--lock=$DENO_LOCK
.DENO_PERMISSIONS
so deno only runs with a specific list of permissions. Deno lambda requires at least--allow-env
and--allow-net
.DENO_PREFIX
prepends to console.log etc. a template literal, this can includerequestId
andlevel
variables only (default${level}\tRequestId: ${requestId}\r
).DENO_UNSTABLE
so deno runs with the--unstable
.
Further configuration TBD.
deno-lambda exports Definitely Typed's aws-lambda types, listed in https://deno.land/x/lambda/mod.ts and defined in https://deno.land/x/lambda/types.d.ts.
It's good practice to reference the trigger's type in the handler, for example:
APIGateway use APIGatewayProxyEventV2
and APIGatewayProxyResultV2
, SQS use
SQSEvent
, etc.
Note: Despite there being multiple interfaces with the Context
suffix, the
second handler argument must be Context
. These other interfaces can be
accessed from the first argument (the event), for example event.requestContext
of an APIGatewayProxyEventV2
.
The recommended way to deploy is to use the
SAR application
and either reference the outputted LayerArn
as a layer in your function.
SAR application (see quick start)
- SAM example
- Serverless example (feat. Dynamodb)
- AWS CDK example
- Docker container example
- Zipped source example,
deno-lambda-example.zip
, see bundling-code section.
See earlier version of quick start for a walkthrough of how to bundle yourself.
See the deno_dir-remapping section for how to include the correct DENO_DIR files to avoid any runtime compilation.
The way the lambda platform works means that promises not awaited in the handler may never be completed. This is because the underlying container can be suspended between invocations and will sometimes be shutdown afterwards.
export async function badHandler(
event: APIGatewayProxyEventV2,
context: Context,
): Promise<APIGatewayProxyResultV2> {
somethingAsync(); // not awaited so may not complete
return { statusCode: 200, body: "" };
}
export async function goodHandler(
event: APIGatewayProxyEventV2,
context: Context,
): Promise<APIGatewayProxyResultV2> {
await somethingAsync();
return { statusCode: 200, body: "" };
}
If you need to return immediately but want to invoke a longer running process
you can
async-invoke
another lambda function (that does the await somethingAsync()
).
- deno support for architect
- deno on zeit now
- A layer for AWS Lambda that allows your functions to use
git
andssh
binaries
Create a zip file which contains:
- an entry point which exports an async function (e.g.
hello.ts
) - include any other files needed to run the entry file
- (optional but preferred) .deno_dir directory
*You can use a different directory with DENO_DIR
environment variable.
Alternatively use deno bundle
command and include the outputted js file, see
also HANDLER_EXT
.
In order for compile artifacts to be recovered (and avoid runtime compilation) you must do the following directory remapping:
# Compile the handler (and cache dependencies and compile artifacts into DENO_DIR).
DENO_DIR=.deno_dir deno cache hello.ts
# This is the "remapping" step:
cp -R .deno_dir/gen/file/$PWD/ .deno_dir/LAMBDA_TASK_ROOT
# Note: We do the inverse of this operation in bootstrap.
zip lambda.zip -x '.deno_dir/gen/file/*' -r .deno_dir hello.ts # other source files
In a serverless.yml
this can be automatically prior to each upload using the
serverless-scriptable-plugin
:
plugins:
- serverless-scriptable-plugin
custom:
scriptHooks:
before:package:createDeploymentArtifacts: DENO_DIR=.deno_dir deno cache api/candidate.ts && cp -R .deno_dir/gen/file/$PWD/ .deno_dir/LAMBDA_TASK_ROOT
See
example-serverless/serverless.yml
.
You can execute deno-lambda locally using docker-lambda. First, unzip the deno-lambda-layer.zip layer into a directory.
Now, from the directory of your application:
# replace LAYER_DIR with the directory you unzipped the layer to e.g. $PWD/layer
# replace hello.handler with your file/handler function
# replace '{}' with the json to pass to the handler
$ docker run -it --rm -v "$PWD":/var/task:ro,delegated -v "LAYER_DIR":/opt:ro,delegated lambci/lambda:provided.al2 hello.handler '{}'
# handler response from goes to stdout
To execute multiple times AKA "stay-open" API mode:
# replace LAYER_DIR with the directory you unzipped to e.g. $PWD/layer
# replace hello.handler with your file.handler function
$ docker run -e DOCKER_LAMBDA_STAY_OPEN=1 -p 9001:9001 -it --rm -v "$PWD":/var/task:ro,delegated -v "LAYER_DIR":/opt:ro,delegated lambci/lambda:provided.al2 hello.handler
Lambda API listening on port 9001...
and in another terminal:
# replace '{}' with the json to pass to the handler
$ aws lambda invoke --endpoint http://localhost:9001 --no-sign-request --function-name deno-func --payload '{}' output.json
# output.json is populated with the handler response
For more advanced usage, e.g. including multiple layers and installing additional libraries (with yumbda), please consult the docker-lambda documentation.
You can set the way console.log
etc. outputs to cloudwatch logs using
DENO_PREFIX
. You can include the line number using
${(new Error).stack.split('\n')[4]}
or the datetime using
${new Date().toISOString()}
(though the datetime of every cloudwatch event is
part of the event itself.
Use these as a template literal, for example:
DENO_PREFIX=${level}\t${requestId}\t${(new Error).stack.split('\n')[4]}\r
This will prefix each log with the level, the request id and the line number:
Many thanks to Yoshiya Hinosawa's blogpost for the initial work on this runtime.