Declarative flush callback

[tpluscode]

It is sometimes useful to take advantage of the flush callback to step streams.

For example, I used the SHACL step in a way that pipeline does not fail immediately on error but stores reports in memory and only fails the pipeline at the end. Another use case that could be implemented with flush would be to call hooks when pipeline finishes

My current SHACL looks somewhat like this

<#validate>
  a :Step ;
  code:implementedBy
    [
      code:link <node:barnard59-validate-shacl/validate.js#shacl> ;
      a code:EcmaScriptModule ;
    ] ;
  code:arguments
    [ code:name "shape" ; code:value <#Shapes> ] ,
    [
      code:name "onViolation" ;
      code:value [ a code:EcmaScriptModule ; code:link <file:../lib/shacl.js#collectShaclReports> ] ;
    ];
.

<#checkShaclReports>
   a :Step ;
  code:implementedBy
    [
      code:link <file:../lib/shacl.js#failOnAnyViolations> ;
      a code:EcmaScriptModule ;
    ] ;
.

The failOnAnyViolations stream is essentially a PassThrough with the second callback.

I would like to propose an optional property for steps, which would combine both:

<#validate>
  a :Step ;
  code:implementedBy
    [
      code:link <node:barnard59-validate-shacl/validate.js#shacl> ;
      a code:EcmaScriptModule ;
    ] ;
  code:arguments
    [ code:name "shape" ; code:value <#Shapes> ] ,
    [
      code:name "onViolation" ;
      code:value [ a code:EcmaScriptModule ; code:link <file:../lib/shacl.js#collectShaclReports> ] ;
    ];
+ :after
+   [
+     code:implementedBy
+       [ 
+         code:link <file:../lib/shacl.js#failOnAnyViolations> ;
+         a code:EcmaScriptModule ;
+       ] ;
+   ] ;
.

-<#checkShaclReports>
-  a :Step ;
- code:implementedBy
-   [
-     code:link <file:../lib/shacl.js#failOnAnyViolations> ;
-     a code:EcmaScriptModule ;
-   ] ;
-.