Amplify-cli: amplify api push produces "Cannot convert undefined or null to object"

[vientang]

Describe the bug
I have an amplify project with an appsync graphql api that was configured through the amplify cli. I want to update my schema. In amplify/backend/api/schema.json, I've added a new field, then ran amplify api push.

Here's the output.
Are you sure you want to continue? Yes

> An error occurred when pushing the resources to the cloud
> Cannot convert undefined or null to object

To Reproduce
Steps to reproduce the behavior:

1. Add a graphql api with amplify add api
2. amplify api push
3. Add a new field to the schema locally in amplify/backend/api/schema.json
4. amplify api push again

Expected behavior
After updating the schema file, I expected the amplify api push command to resolve successfully and update the schema with the new field

Additional context

• The schema has a @model directive (by default with amplify add api)
• DynamoDB tables were created. Data was saved successfully. (I manually removed the entries in the table and had the same issue)
• Another dev had configured his machine with this project (not with multienv)
• I tried amplify configure project and received the same error. This time with a trace stack.
• 
  Could this might be related?
• @aws-amplify/[email protected]
  
  
  
  • React project using Gatsby
  
  

Desktop (please complete the following information):

• OS: macOS 10.13.6
• Browser: Chrome
• Version: 72.0.3626.96

[aada]

This happens also after I create custom stack in stacks folder
stacks/customstack.json :

{
    "AWSTemplateFormatVersion": "2010-09-09",
    "Resources": {
        "S3Bucket": {
            "Type": "AWS::S3::Bucket"
        }
    }
}

@vientang can you check if this happens to you

note : this issue doesn't happen when I update schema.graphql
Desktop

• OS: macOS 10.14
• Amplify Cli : 1.1.2

[vientang]

@aada Interesting, this actually might be triggering the "Cannot convert undefined or null to object" message.

1. I've reverted my project back, with the scheme.graphql file left unchanged.
2. Run amplify push

I'm still getting this error message. I didn't create a custom stack but somehow when I run amplify status, it shows the hosting category operation as Update. Could this be because another developer had configured this project on their machine?

[aada]

@vientang I fixed the issue by adding Parameters : {} to my custom stack
I think the problem was when amplify validates cloudformation templates it passes the parameters defined in ./amplify/backend/api/{your api name}/parameters.json and that becomes invalid when there are no parameters in the template.

It would very helpful if amplif cli had verbose option
I hope this helps you

Could this be because another developer had configured this project on their machine?

my teammates had some problems when multiple developers worked on an amplify project.
let me check how did they resolve it.

[aada]

this article might be helpful multi-environments

[vientang]

This got me a little bit further. I've added the Parameters: {} to ./amplify/backend/api/{your api name}/parameters.json.

Now, I get the message "GraphQL schema compiled successfully. Edit your schema at ... " but right after, "An error occurred when pushing the resources to the cloud Cannot convert undefined or null to object".

I'll look into the multi-environments link you've sent. Thanks!

[vientang]

What I ended up doing was starting over, removing all of the services and adding them back in. This time, when I initialized the project, the project-config.json file is different.

Before

"providers": [
        "awscloudformation"
    ],
"frontend": "javascript",

Now

"frontendHandler": {
        "javascript": "/usr/local/lib/node_modules/@aws-amplify/cli/node_modules/amplify-frontend-javascript"
    },
"providers": {
        "awscloudformation": "/usr/local/lib/node_modules/@aws-amplify/cli/node_modules/amplify-provider-awscloudformation"
    },
"projectPath": "/path/to/my/project",

I'm unsure if this has anything to do with the issue, but updating the schema works just as described in the documentation. Thanks for looking into it @aada!

[aada]

I think the problem happens due to updating amplify cli. The new versions have different structure. Can you check if the creator of your project has an older version of amplfy cli?

[kaustavghosh06]

Hey guys, could you install the latest version of the CLI and see if the issue still exists? In the newer version we don't mantain absolute paths for the plugins (and we do a auto-migration) for you as well when upgrading to the latest version of the CLI. Please take a look at this accouncement - https://aws.amazon.com/blogs/mobile/amplify-adds-support-for-multiple-environments-custom-resolvers-larger-data-models-and-iam-roles-including-mfa/

[vientang]

@kaustavghosh06 Sure, I’ll take a look at the new cli. My current version is 1.1.2. The true test will be in on boarding another developer. Looks like the new env command would help (and making sure we’re using the same version of amplify cli).

I think this ticket can be closed. After restarting my project, I was able to successfully initialize and update the GraphQL schema. I did the update (amplify push) many times just to be sure.

@aada I’m waiting for his response but 99% sure his version is different from my previous and new version of amplify cli.

Here’s some context in case it helps. First time I initialized the project was in mid December. Roughly 3 weeks later, the other developer configured the project on his machine. We didn’t compare our folder structure then but everything was working - auth via Cognito, hosting via S3 and reading/writing to AppSync. This issue came up when I tried to update the api. The other developer did not have issues updating the api on his machine.

[kaustavghosh06]

@vientang Closing this issue as you mentioned is not an issue for you and your team anymore. please let me know if you're still seeing this issue and we can re-open this.

[kevcam4891]

I just found success in running amplify env checkout [your-environment-name]. I made sure to git commit first just to ensure nothing in progress got lost.

Also just noticed that I had a build file open in my editor (VSCode). The User.json file in this case. Might have been no factor, but I am not certain what sort of locks the editor puts on the build files that might conflict with the CLI during the compile process.