Re: [tosca] Complex property composition

[Tal Liron <tliron@redhat.com>]

You are absolutely right about the ambiguity. Please see the discussion on this mailing list called "New syntax for function calls" that attempts to resolve this for TOSCA 2.0.

As for TOSCA 1.X, it was not specified but implied that a parser would prioritize detecting a function call first (a map with a single string key which is one of the recognized function names). But it was very much left up to individual implementations to decide what to do. Would you get an error telling you that a function name is not known? Or would it be simply treated as a map value, leading to a data error if the data type is not a map?

(And there is also a rare edge case in which you actually have a data type property that is named the same as a TOSCA function, or you want to use such a name as a map key. TOSCA 1.X would simply not let you do this because it would interpret that notation as a function call.)

Your point about nesting function calls as arguments to other functions is also on point. The TOSCA 1.X specs really did not clarify this. However, in examples we did see use of nesting. Again, I hope you can contribute to the "New syntax for function calls" discussion, which goes into this in some detail.

To echo what Chris said, even when the spec is not entirely clear we can still have a good understanding of the "spirit" of what was intended. There are many such problems with the TOSCA 1.X spec. We are working hard to resolve these ambiguities in TOSCA 2.0. Your feedback here is important. Whatever we do, we absolutely must clarify how nesting works in a deterministic way.

On Wed, Apr 21, 2021 at 1:40 PM Nemeth, Denes (Nokia - HU/Budapest) <denes.nemeth@nokia.com> wrote:

Dear Tal

Â

Are you sure about this, because the spec says that :

â3.6.11.2 Grammar

Property assignments have the following grammar:

3.6.11.2.1 Short notation:

The following single-line grammar may be used when a simple value assignment is needed:

<property_name>: <property_value> | { <property_value_expression> }

In the above grammars, the pseudo values that appear in angle brackets have the following meaning:

ÂÂÂÂÂÂÂÂÂÂproperty_name:Ârepresents the name of a property that would be used to select a property definition with the same name within on a TOSCA entity (e.g., Node Template, Relationship Template, etc.,) which is declared in its declared type (e.g., a Node Type, Node Template, Capability Type, etc.).ÂÂ

ÂÂÂÂÂÂÂÂÂÂproperty_value,Âproperty_value_expression:Ârepresent the type-compatible value to assign to the named property. Property values may be provided as the result from the evaluation of an _expression_ or a function.

â

Â

For me this definition would mean that the property can either have a fixed value assignment or can be specified via a Tosca function. It does not say anywhere that a property may be constructed from multiple functions.

Would not the possibility of constructing a complex property look like this?

Â

<propertyName>: <property_value> | { <property_value_expression> } | <composite>

<composite>:

ÂÂ <key>: <property_value> | <composite>

, where <key> is a fixed value or { <property_value_expression> }

Â

Â

Maybe Puccini has more advanced capabilities that is described in the specs.

Â

My second problem with this approach is that the parser can not the value of the property

p1:

ÂÂchild1:

ÂÂÂÂget_input: simple_input_child1

value of the child1 is

• get_input: simple_input_child (so a map that has one element with get_input key and simple_input_child value

xor

• The value of the get_input function

In this case the parser is lucky because it can deduct that it is the 2nd Case (because of the type of child1 is string).

However in a generic case when such deduction is not possible and a complex property is constructed from N number of Tosca functions or fixed values.

The complex value may have 2^N variants (exponential), how should we choose which is the actual value.

Here is an example for the problem.

tosca_definitions_version: tosca_simple_yaml_1_2

data_types:

 data_type_1:

ÂÂÂ derived_from: tosca.datatypes.Root

ÂÂÂ properties:

ÂÂÂÂÂ child1:

ÂÂÂÂÂÂÂ type: map

ÂÂÂÂÂ child2:

ÂÂÂÂÂÂÂ type: map

node_types:

 node_type_1:

ÂÂÂ properties:

ÂÂÂÂÂ p1:

ÂÂÂÂÂÂÂ type: data_type_1

topology_template:

 inputs:

ÂÂÂ simple_input_child1:

ÂÂÂÂÂ type: map

ÂÂÂÂÂ default:

ÂÂÂÂÂÂÂ a: b

ÂÂÂ simple_input_child2:

ÂÂÂÂÂ type: map

ÂÂÂÂÂ default:

ÂÂÂÂÂÂÂ c: d

 node_templates:

ÂÂÂ node_template_2:

ÂÂÂÂÂ type: node_type_1

ÂÂÂÂÂ properties:

ÂÂÂÂÂÂÂ p1:

ÂÂÂÂÂÂÂÂÂ child1:

ÂÂÂÂÂÂ ÂÂÂÂÂget_input: simple_input_child1

ÂÂÂÂÂÂÂÂÂ child2:

ÂÂÂÂÂÂÂÂÂÂÂ get_input: simple_input_child2

Â

Â

What is the resolved value of p1:

Â

child1:

 a: b

child2:

 c: d

Â

or

Â

child1:

 get_input: simple_input_child1

child2:

 c: d

Â

or

Â

child1:

 get_input: simple_input_child2

child2:

 c: d

Â

or

Â

child1:

 get_input: simple_input_child1

child2:

 get_input: simple_input_child2

Â

I think that at least some kind of precedence would be required between values and function. For example if the value of a property can be evaluated as a Tosca function, than it is considered to be the value of the tosca function.

Â

From: tosca@lists.oasis-open.org <tosca@lists.oasis-open.org> on behalf of Tal Liron <tliron@redhat.com>
Date: Wednesday, 2021. April 21. 18:26
To: Nemeth, Denes (Nokia - HU/Budapest) <denes.nemeth@nokia.com>
Cc: tosca@lists.oasis-open.org <tosca@lists.oasis-open.org>
Subject: Re: [tosca] Complex property composition

Yes, I'm sure it is possible.

You can think of it this way: a data type just like a node template also has properties, so these both accept assigned values in the same way. This implies that you can nest values to any depth.

The situation is less clear when we are dealing with list and map types. However, I think that it likewise should be possible. And for maps, it should even be possible to use function calls for map keys. (Puccini supports all these cases.) Example:

node_types:
 DataNode:
  properties:
   string_map:
    type: map
    entry_schema: string
topology_template:
 node_templates:
  data:
   type: DataNode
   properties:
    string_map:
     Greeting: Hello
     Message: { concat: [ Good, ' ', Day ] }
     { concat: [ Recip, ient ] }: Puccini

Â

On Wed, Apr 21, 2021 at 9:50 AM Nemeth, Denes (Nokia - HU/Budapest) <denes.nemeth@nokia.com> wrote:

Dear Tosca community

Â

Can you help us to clarify the usage of property assigment in node templates specified in

https://docs.oasis-open.org/tosca/TOSCA-Simple-Profile-YAML/v1.3/os/TOSCA-Simple-Profile-YAML-v1.3-os.html#DEFN_ELEMENT_PROPERTY_VALUE_ASSIGNMENT

Â

Is is possible to assign parts of complex properties to node templates using property value expressions or not?

Â

Node_template_1 assigns complex_input input to p1 property. There is several examples to this in the specs.

Â

Node_template_2 constucts the value of p1 from multiple input. Is this valid or not?

Â

Many thanks, for the help

Â

tosca_definitions_version: tosca_simple_yaml_1_2

data_types:

 data_type_1:

ÂÂÂ derived_from: tosca.datatypes.Root

ÂÂÂ properties:

ÂÂÂÂÂ child1:

ÂÂÂÂÂÂÂ type: string

ÂÂÂÂÂ child2:

ÂÂÂÂÂÂÂ type: string

node_types:

 node_type_1:

ÂÂÂ properties:

ÂÂÂÂÂ p1:

ÂÂÂÂÂÂÂ type: data_type_1

topology_template:

 inputs:

ÂÂÂ complex_input:

ÂÂÂÂÂ type: data_type_1

ÂÂÂ simple_input_child1:

ÂÂÂÂÂ type: string

ÂÂÂ simple_input_child2:

ÂÂÂÂÂ type: string

 node_templates:

ÂÂÂ node_template_1:

ÂÂÂÂÂ type: node_type_1

ÂÂÂÂÂ properties:

ÂÂÂÂÂÂÂ p1: { get_input: complex_input }

ÂÂÂ node_template_2:

ÂÂÂÂÂ type: node_type_1

ÂÂÂÂÂ properties:

ÂÂÂÂÂÂÂ p1:

ÂÂÂÂÂÂÂÂÂ child1: { get_input: simple_input_child1 }

ÂÂÂÂÂÂÂÂÂ child2: { get_input: simple_input_child2 }