Allow templates to define complex mappings

[arthurdm]

This was brought up in #25 (review)

We should have a way to define complex mappings inside an annotation / x-descriptor.

[sbose78]

cc @Avni-Sharma

[navidsh]

@navidsh could you help me with a complete example please? I will work on introducing this in a separate PR today.

@sbose78 Example for template mapping (re #25 (comment)):

Kafka CR:

status:
  listeners:
    - type: plain
      addresses:
        - host: my-cluster-kafka-bootstrap.service-binding-demo.svc
          port: 9092
        - host: my-cluster-kafka-bootstrap.service-binding-demo.svc
          port: 9093
    - type: tls
      addresses:
        - host: my-cluster-kafka-bootstrap.service-binding-demo.svc
          port: 9094

Descriptors:

    - path: listeners
      x-descriptors:
        - servicebinding:prefix:elementType=template:source:{{ range $idx1, $el1 := .status.listeners }}{{ range $idx2, $el2 := $el1.addresses }}{{ $el1.type }}_{{ $idx2 }}={{ $el2.host }}:{{ $el2.port }}{{ printf "\n" }}{{ end }}{{ end }}:bindAs=envVar

Same template from the descriptor but pretty formatted:

{{- range $idx1, $el1 := .status.listeners -}}
  {{- range $idx2, $el2 := $el1.addresses -}}
    {{ $el1.type }}_{{ $idx2 }}={{ $el2.host }}:{{ $el2.port }}{{printf "\n" }}
  {{- end -}}
{{- end -}}

Executing the template on the Kafka CR example above produces this:

plain_0=my-cluster-kafka-bootstrap.service-binding-demo.svc:9092
plain_1=my-cluster-kafka-bootstrap.service-binding-demo.svc:9093
tls_0=my-cluster-kafka-bootstrap.service-binding-demo.svc:9094

Still it needs some thinking about how properly distinguish keys from values.
<key>=<value> doesn't quite work since <value> may have = in it. It needs a better delimiter to define the boundaries of key-value pairs. (But I'm out of idea at this time of day. lol)

[arthurdm]

@sbose78 - would you want to take a stab at this one?

[sbose78]

Still it needs some thinking about how properly distinguish keys from values.

@navidsh , the key should never have the = in it, so we always split on the first =?

[navidsh]

@sbose78 The template in my example, produces the following output as "one" string once it executes:

plain_0=my-cluster-kafka-bootstrap.service-binding-demo.svc:9092
plain_1=my-cluster-kafka-bootstrap.service-binding-demo.svc:9093
tls_0=my-cluster-kafka-bootstrap.service-binding-demo.svc:9094

SBO should then, take this text and turn it into a map e.g.

{
  "plain_0": "my-cluster-kafka-bootstrap.service-binding-demo.svc:9092",
  "plain_1": "my-cluster-kafka-bootstrap.service-binding-demo.svc:9093",
  "tls_0": "my-cluster-kafka-bootstrap.service-binding-demo.svc:9094"
}

It's true that the key cannot have a = but the value can include = and \n which makes it hard to differentiate where a value ends and a key starts. The template is relying on = and \n as delimiters.

Maybe picking better delimiters or a better structure could make it easier to convert the output text into key-value map.

[navidsh]

Kafka CR:

status:
  listeners:
    - type: plain
      addresses:
        - host: my-cluster-kafka-bootstrap.service-binding-demo.svc
          port: 9092
        - host: my-cluster-kafka-bootstrap.service-binding-demo.svc
          port: 9093
    - type: tls
      addresses:
        - host: my-cluster-kafka-bootstrap.service-binding-demo.svc
          port: 9094

Descriptors:

    - path: listeners
      x-descriptors:
        - servicebinding:elementType=template:source={{ GO TEMPLATE }}:bindAs=envVar

Go Template:

{{- range $idx1, $lis := .status.listeners -}}
  {{- range $idx2, $adr := $el1.addresses -}}
    {{ $lis.type }}_{{ $idx2 }}={{ printf "%s:%s\n" "$adr.host" "$adr.port" | b64enc | quote }}
  {{- end -}}
{{- end -}}

This template produces as one string:

plain_0="<base64 of my-cluster-kafka-bootstrap.service-binding-demo.svc:9092>"
plain_1="<base64 of my-cluster-kafka-bootstrap.service-binding-demo.svc:9093>"
tls_0="<base64 my-cluster-kafka-bootstrap.service-binding-demo.svc:9094>"

The operator would then parse the above string into secret key-value pairs.

[sbose78]

I'm good with this. Let's take this in.