This provides a sample of how to write a CDP Enhanced Power Platform Connector.

Documentation for protocol is here: https://aka.ms/PowerFxCDP

• CdpSampleWebAPI - a REST ASP to impl CDP. Runs as an ASP.Net WebAPI.

• CdpValidator - a client-side API that can be pointed at a CDP endpoint to validate it. This can point to existing connectors (hosted in APIM) or to localhost (such as locally running CdpSampleWebAPI)

This is implemented on top of Power Fx definitions, notably RecordType. These are C# classes that already describe tabular operations:

https://github.com/microsoft/Power-Fx/blob/main/src/libraries/Microsoft.PowerFx.Core/Public/Types/RecordType.cs

You can clone this repo, and override ITableProvider to provide your datasource. This project will then plumb between RecordType and CDP protocol.

This is a command line tool which is using the following options

• -auth <path to Json file>
• -logdir <path to log directory>

When run from Visual Studio, those parameters can be set in Properties\launchSettings.json file

For running the validation with an APIM/CDP connector

{
  "endpoint":      "<APIM endpoint>",
  "environmentId": "<Environment Id>",
  "connectionId":  "<Connection Id>",
  "urlprefix":     "/apim/<Connector name>/{connectionId}",
  "dataset":       "<Dataset>",
  "tablename":     "<Table>",
  "jwtFile":       "<Path to JWT txt file>"
}

Here is an example (remove "connectionId" parameter when testing non-APIM endpoint during devlopment)

{
  "endpoint":      "49970107-0806-e5a7-be5e-7c60e2750f01.12.common.firstrelease.azure-apihub.net",
  "environmentId": "49970107-0806-e5a7-be5e-7c60e2750f01",
  "connectionId":  "4ac3586f00be475a8fcd09e65128c9ca",
  "urlprefix":     "/apim/sharepointonline/{connectionId}",
  "dataset":       "https://microsoft.sharepoint.com/sites/MySite",
  "tablename":     "Documents",
  "jwtFile":       "C:\\temp\\jwt.txt"
}

For selecting multiple datasets, you can use a star.
For selecting multiple tables, you can use a star.

By default, CDP validator will define a glocal timeout of 15 minutes for the entire set of checks. This is not configurable on the command line but can be modified in Program.cs file.

Initial checks

• Get dataset metadata (/$metadata.json/datasets)
• List of datasets (/datasets)

For each dataset (except if the dataset is specified in the config file, where only one dataset will be tested)

• Get list of tables (/datasets/{datasetName}/tables)

For each table in the dataset[s] (except if the table is specified in the config file, where only onle table will be tested)

• Get table schema (/$metadata.json/datasets/{datasetName}/tables/{tableName})
• Validate each field type (cannot be Untyped Object or missing/undefined)
• Retrieve one row with $top and verify we only get one row (or zero)
• Retrieve 100 rows with $top and verify we get at most 100 rows
• Retrieve 5000 rows with $top and expect we will get less than 2100 rows (we expect paging to take place)

For tables having 2 or more columns, following tests are executed otherwise a warning is generated (and those tests skipped)

• Retrieve 10 rows (with $top) and selecting 1 column ($select)
  The column that is used is the first having a String or Number or Decimal or Guid or DateTime or Boolean type
• Retrieve 10 rows (with $top) and selecting 2 columns ($select), using the same column selection logic
  

For tables with 5 or more rows, following tests are executed otherwise a warning is generated (and those tests skipped)

• Using eq, ne, gt, ge, lt, le operators (6 different queries), retrieve 10 rows using $filter with 1 field and using a value of the 1st row for that column
• Using eq, ne, gt, ge, lt, le operators (6 different queries), retrieve 10 rows using $filter with 2 fields and using a value of the 1st row for these columns
• Try retrieving 10 rows with an invalid $filter syntax ($filter=<space>)
• Retrieve 10 rows using $orderby with one column
• Retrieve 10 rows using $orderby with 3 columns

Finally

• Try getting the schema of an invalid table name (random Guid) and expecting a 400 or 404 error

In the log folder, you'll find

• one YAML file per network call containing the request, response status code and content if any
• an Errors.http file (which can be opened in VS code) and reporting all failues
  These failures can be replayed for troubleshooting