lambda ::

[Update: Apparently you can get there from here! That is, if you use firebase-admin instead of @google-cloud/firestore. I'll have more on this in the future, but the gist of it is summarized here.]

[ 更新：显然您可以从这里到达那里！也就是说，如果您使用firebase-admin而不是@google-cloud/firestore 。我将有更多的关于这方面的未来，但它的要点总结在这里。]

A while back I was exploring Netlify's support for FaunaDB: a NoSQL document-oriented database with some special features to handle transactions across dispersed database servers. I decided to try it because it was a convenient choice, since there was example code I could start with. The example used lambda functions as a frontend to the database.

前一段时间，我正在探索Netlify对FaunaDB的支持：NoSQL面向文档的数据库，具有一些特殊功能，可以处理分散的数据库服务器之间的事务。我决定尝试一下，因为这是一个方便的选择，因为我可以从示例代码开始。该示例使用lambda函数作为数据库的前端。

I modified the original lambda functions to talk to the FaunaDB GraphQL API (instead of FQL). While that worked, in the end I felt Fauna's GraphQL support wasn't quite ripe yet, so I looked around for alternatives.

我修改了原始的lambda函数，以便与FaunaDB GraphQL API(而不是FQL )进行通信。在此过程中，最后，我感到Fauna对GraphQL的支持还不太成熟，因此我四处寻找替代方案。

Eventually I settled on Cloud Firestore. I based this new project on the Fauna example, swapping out the faunadb module with apollo-server-lambda, so that I could write my own GraphQL API and resolvers.

最后，我选择了Cloud Firestore 。我基于对动物例如这个新项目，换出与faunadb模块阿波罗-服务器-拉姆达，这样我就可以写我自己的GraphQL API和解析器。

One of the refinements I had to make was to push all my Netlify Function dependencies down to the /functions folder in my project (separate and at the same level as the /src folder that contains my React client). To do this, I ran npm init while inside the functions folder, moved a set of dependencies from the top-level package.json to the new /functions/package.json, added a webpack.functions.js, then ran yarn install to pull the packages into a new node_modules folder.

我必须做的一项改进是将我所有的Netlify Function依赖项下推到项目中的/ functions文件夹(与包含我的React客户端的/ src文件夹位于同一级别)。为此，我在functions文件夹内运行了npm init ，将一组依赖项从顶级package.json移至新的/functions/package.json，添加了webpack.functions.js ，然后将yarn install运行到将软件包拉到新的node_modules文件夹中。

The result was this:

结果是这样的：

I'll talk about the subfolders later; the main thing to notice is that there's yarn files, plus package.json, a node_modules folder, a schema folder, and some .js files for testing.

稍后我将讨论子文件夹。最要注意的是，这里有yarn文件，package.json，node_modules文件夹，schema文件夹以及一些用于测试的.js文件。

The original project used netlify_lambda to build, which uses webpack and babel. I ran into some issues, fixed them, then ran into them again later.

原始项目使用netlify_lambda进行构建，该项目使用webpack和babel。我遇到了一些问题，将其修复，然后稍后再次遇到它们。

Frustrated, I decided to forego netlify-lambda and chose Netlify Dev to build and deploy from the command line. The drawback was that I didn't have the ability to launch a local server, but I could deploy candidates to Netlify and test them without first checking source into github or deploying directly to production.

沮丧的是，我决定放弃netlify-lambda，选择了Netlify Dev从命令行进行构建和部署。缺点是我没有启动本地服务器的能力，但是我可以将候选人部署到Netlify并对其进行测试，而无需先将源检查到github或直接部署到生产中。

There were less moving parts since webpack and babel were no longer needed. When going this route, you probably set the environment variable AWS_LAMBDA_JS_RUNTIME to nodejs10.x in the Build & deploy settings for your functions.

由于不再需要webpack和babel，因此移动部件更少。在执行此路由时，您可能在函数的“ 构建和部署”设置中将环境变量AWS_LAMBDA_JS_RUNTIME设置为nodejs10.x 。

事情并不总是像看起来那样 (Things are not always as they seem)

More familiar with GraphQL clients and servers than with lambda functions in the cloud, I had some naive assumptions about how things got deployed in Netlify. I thought functions were more or less copied over and build scripts run on the server, where all would be happy and my functions would be callable via URLs.

我对GraphQL客户端和服务器比对云中的lambda函数更加熟悉，我对如何在Netlify中部署事物有一些幼稚的假设。我认为函数或多或少被复制了，并且构建脚本在服务器上运行，在那里一切都会很高兴，并且我的函数可以通过URL进行调用。

This is not at all what happens.

这根本不发生什么。

When I started with netlify_lambda, it would use webpack to create a functions_build output file. My netlify.toml configuration had that as the functions location.

当我从netlify_lambda开始时，它将使用webpack创建一个functions_build输出文件。我的netlify.toml配置将其作为函数位置。

[build]functions = "functions-build"# This will be run the site buildcommand = "yarn build"# This is the directory is publishing to netlify's CDNpublish = "build"

When I switch to using Netlify Dev, I dispensed with the output folder and just deployed the "unbundled" /functions source. That's not the end of the story, though.

当我切换为使用Netlify Dev时，我放弃了输出文件夹，而只是部署了“未捆绑” / 功能源。不过，这还不是故事的结局。

身份验证问题 (Authentication woes)

In the FaunaDB project, authentication was through an environment variable whose value was a simple token. A similar mechanism is used by Firebase, but instead of a token, the variable value is a path to a credentials file that you generate through the FireBase console. The lambda functions create a Firebase instance, and that instance looks for the env variable to locate the credentials file for authentication.

在FaunaDB项目中，身份验证是通过环境变量进行的，其值是一个简单的令牌。 Firebase使用类似的机制，但是变量值代替令牌，是通过FireBase控制台生成的凭证文件的路径。 lambda函数创建一个Firebase实例，该实例查找env变量以找到用于身份验证的凭据文件。

It seems like no matter where I put that credentials file or what path I used, the Firebase client would fail to find it. In the course of my research I came across a mention of Netlify's zip-it-and-ship-it utility, which other people with other problems recommended for bundling up functions in zip files.

无论我将凭据文件放在哪里或使用什么路径，Firebase客户端似乎都找不到它。在研究过程中，我提到了Netlify的zip-it-and-ship-it实用程序，建议其他有其他问题的人将zip文件的功能捆绑在一起。

I tried it, modifying the build process to call a NodeJS script that zipped up my functions to a functions-dist folder (changing the netlify.toml config to no point to that instead of the functions source folder). Although it didn't immediately fix my issues with the credentials file, I noticed some things.

我尝试了一下，修改了构建过程，以调用一个NodeJS脚本，该脚本将我的函数压缩到functions-dist文件夹中(将netlify.toml配置更改为指向该文件夹而不是函数 source文件夹)。尽管它不能立即解决凭据文件的问题，但我注意到了一些问题。

I began to realize that as each lambda function .js file was bundled up into a zip file, it also contained its own node_modules folder. What's more, the node_modules folder was "customized" to contain only those dependencies explicitly required by each function.

我开始意识到，由于每个lambda函数.js文件都捆绑到一个zip文件中，因此它还包含自己的node_modules文件夹。此外，“定制” node_modules文件夹以仅包含每个函数明确要求的那些依赖项。

聪明，但不够聪明 (Clever, but not clever enough)

It took some thinking, but I decided that if I added my .json file in a local project, then made it a dependency to each lambda function, it would be pulled in the node_modules folder. At that point, I would have a path: ./creds/mycred.json. Yay!

它花了一些时间，但我决定，如果我在本地项目中添加.json文件，然后使其成为每个lambda函数的依赖项，它将被拉到node_modules文件夹中。到那时，我将有一条路： ./creds/mycred.json 。好极了！

It didn't quite work--when I examined the zip files, the credential files were there in each zip archive, but the Firebase client still couldn't get to them.

它不是很有效-当我检查zip文件时，每个zip存档中都存在凭证文件，但是Firebase客户端仍然无法访问它们。

I confessed my utter failure on the Netlify support forum, saying that I planned to join a commune to learn to weave hammocks.

我在Netlify支持论坛上承认自己完全失败，并说我计划加入一个公社来学习编织吊床。

救命！ (Help!)

I must have evoked some pity, as Dennis from Netlify soon responded and let me know that lambda functions cannot actually access the file system. What I was attempting (loading credentials via a file path) was impossible. He suggested importing the file into each lambda .js (which I had already done). It doesn't appear, though, that the Firebase client allows you to pull in credentials via an import.

我一定引起了一些同情，因为Netlify的Dennis很快做出了回应，并让我知道lambda函数实际上无法访问文件系统。我尝试的事情(通过文件路径加载凭据)是不可能的。他建议将文件导入每个lambda .js(我已经完成了)。但是，似乎没有Firebase客户端允许您通过导入拉入凭据。

That aside, Dennis sort of hinted that perhaps this isn't really the approach I should take, anyway. He had a point. The only reason I went this route was because I was following one of Netlify's examples, but swapping out the faunadb package with apollo-server-lambda might just have added a lot more weight to the lambda functions; if so, it would likely have an affect on spin-up times during cold starts.

除此之外，丹尼斯有点暗示也许这不是我应该采取的方法。他有一点。我走这条路线的唯一原因是因为我遵循的是Netlify的示例之一，但是用apollo-server-lambda换出Animaldb程序包可能只会增加lambda函数的权重。如果是这样，则可能会影响冷启动期间的加速时间。

抛弃Lambda函数 (Ditching lambda functions)

Lambda functions are not a solution for everything. In my case, I only wanted a simple datastore with a GraphQL frontend, without exposing the GraphQL queries in the browser console.

Lambda函数并不能解决所有问题。就我而言，我只想要一个带有GraphQL前端的简单数据存储，而没有在浏览器控制台中公开GraphQL查询。

I can achieve the same ends by having a Node process host both a React client and a GraphQL server. I'm (almost) certain I won't run into any file system access problems, and if so, I'll switch to another method of authentication.

我可以通过让Node进程同时托管React客户端和GraphQL服务器来达到相同的目的。我(几乎)肯定不会遇到任何文件系统访问问题，如果是这样，我将切换到另一种身份验证方法。