Gatsby: Nodes from on-create-node fail to persist after first run/cache

[DSchau]

Description

Using the onCreateNode API works great. However, on subsequent builds (e.g. with a .cache folder persisted), the nodes created from that API do not seem to be cached correctly or otherwise persist, and thus the build breaks.

Steps to reproduce

Demo

In particular, check out the gatsby-node.js file, where I create a custom node based on the contents of a JS file.

Following steps serve to reproduce:

1. Run yarn (or npm install) to install deps
2. Run yarn build and notice that it passes
3. Run yarn build again, and notice the bellow failure:

error GraphQL Error Unknown field `allExampleCode` on type `Query`

  file: /Users/dschau/Code/Work/gatsby/test/on-create-node-repro/src/pages/index.js

   1 | 
   2 |   query {
   3 |     site {
   4 |       siteMetadata {
   5 |         title
   6 |         description
   7 |       }
   8 |     }
   9 | 
> 10 |     examples:allExampleCode {
     |              ^
  11 |       edges {
  12 |         node {
  13 |           code
  14 |           mdAbsolutePath
  15 |         }
  16 |       }
  17 |     }
  18 |     markdown:allMarkdownRemark {
  19 |       edges {
  20 |         node {

Expected result

Build should reliably pass and cache custom node types _or_ they should be created each time (preferably the former).

Actual result

Builds do not reliably pass.

Environment

  System:
    OS: macOS 10.14
    CPU: x64 Intel(R) Core(TM) i9-8950HK CPU @ 2.90GHz
    Shell: 5.3 - /bin/zsh
  Binaries:
    Node: 10.11.0 - /usr/local/bin/node
    Yarn: 1.10.1 - /usr/local/bin/yarn
    npm: 6.4.1 - /usr/local/bin/npm
  Browsers:
    Chrome: 69.0.3497.100
    Firefox: 62.0.2
    Safari: 12.0
  npmPackages:
    gatsby: ^2.0.0 => 2.0.12 
    gatsby-plugin-feed: ^2.0.5 => 2.0.5 
    gatsby-plugin-google-analytics: ^2.0.5 => 2.0.6 
    gatsby-plugin-manifest: ^2.0.2 => 2.0.4 
    gatsby-plugin-offline: ^2.0.5 => 2.0.5 
    gatsby-plugin-react-helmet: ^3.0.0 => 3.0.0 
    gatsby-plugin-sharp: ^2.0.5 => 2.0.5 
    gatsby-plugin-typography: ^2.2.0 => 2.2.0 
    gatsby-remark-copy-linked-files: ^2.0.5 => 2.0.5 
    gatsby-remark-images: ^2.0.1 => 2.0.3 
    gatsby-remark-prismjs: ^3.0.0 => 3.0.1 
    gatsby-remark-responsive-iframe: ^2.0.5 => 2.0.5 
    gatsby-remark-smartypants: ^2.0.5 => 2.0.5 
    gatsby-source-filesystem: ^2.0.1 => 2.0.1 
    gatsby-transformer-remark: ^2.1.1 => 2.1.4 
    gatsby-transformer-sharp: ^2.1.1 => 2.1.3 
  npmGlobalPackages:
    gatsby-dev-cli: 2.4.3
    gatsby: 2.0.0

[DSchau]

For posterity, the issue is that nodes are garbage collected based on the existence of a parent (thanks @KyleAMathews)

So the fix is do something like this (at the plugin level):

createNode({
  id: name,
  children: [],
  parent: node.id,
  code,
  mdAbsolutePath: absolutePath.replace(/\.js$/, '.md'),
  internal: {
    type: 'ExampleCode',
    contentDigest: createContentDigest(JSON.stringify(code)),
  },
});

instead of

createNode({
  id: name,
  children: [],
  parent: `SOME_PARENT_STR`,
  code,
  mdAbsolutePath: absolutePath.replace(/\.js$/, '.md'),
  internal: {
    type: 'ExampleCode',
    contentDigest: createContentDigest(JSON.stringify(code)),
  },
});

[pieh]

We probably could show warning if createNode is used in onCreateNode and it doesn't set passed node as parent?