Declarative programming winning examples

Lightning talk for PyWeb #40

Transcript

1. Declarative
   Programming
   winning examples
   Udi h Bauman, CloudLock
   

   View Slide

2. Declarative
   Programming
   winning examples
   Udi h Bauman, CloudLock
   We’re hiring,
   check us out!
   

   View Slide

3. Declarative Programming is
   a form of high-level programming:
   

   View Slide

4. Declarative Programming is
   a form of high-level programming:
   telling your software what to do, instead of
   how to do it
   

   View Slide

5. Some real-life patterns
   1. “what”-metadata driven
   2. high-level “what”-code
   3. “how”-code abstracted away
   

   View Slide

6. 1. “what”-metadata driven
   ● problem: you have many behaviours that only differ in
   parameters or small variants of behaviour
   

   View Slide

7. 1. “what”-metadata driven
   ● problem: you have many behaviours that only differ in
   parameters or small variants of behaviour
   ● solution: write metadata & code that derives behaviour
   from metadata
   

   View Slide

8. 1. “what”-metadata driven
   ● problem: you have many behaviours that only differ in
   parameters or small variants of behaviour
   ● solution: write metadata & code that derives behaviour
   from metadata
   ● benefit: adding new behaviours now means just adding
   metadata, & in rare cases also overriding
   classes/functions
   

   View Slide

9. Example: rollups aggregation
   

   View Slide

10. Example: rollups aggregation
    

    View Slide

11. 2. high-level “what”-code
    ●
    problem: you have many behaviours that
    only differ in the “edges”
    

    View Slide

12. 2. high-level “what”-code
    ●
    problem: you have many behaviours that
    only differ in the “edges”
    ●
    old solution: using inheritance tricks such
    as template method &c
    

    View Slide

13. 2. high-level “what”-code
    ●
    problem: you have many behaviours that
    only differ in the “edges”
    ●
    newer solution: abstract the algorithm to
    declarative high-level code & “chain” the
    overriding behaviour
    

    View Slide

14. 2. high-level “what”-code
    ●
    problem: you have many behaviours that
    only differ in the “edges”
    ●
    newer solution: abstract the algorithm to
    declarative high-level code & “chain” the
    overriding behaviour
    ●
    benefit: see the big picture & just the
    relevant details
    

    View Slide

15. Example: d3: data-driven documents
    

    View Slide

16. Example: d3: data-driven documents
    

    View Slide

17. 3. “how”-code abstracted away
    ● problem: if some code addresses
    requirements that are obvious & not specific
    to your app, why should it be there?
    

    View Slide

18. 3. “how”-code abstracted away
    ● problem: if some code addresses
    requirements that are obvious & not specific
    to your app, why should it be there?
    ● solution: abstract that code in an underlying
    framework, either yours or someone’s else
    

    View Slide

19. 3. “how”-code abstracted away
    ● problem: if some code addresses
    requirements that are obvious & not specific
    to your app, why should it be there?
    ● solution: abstract that code in an underlying
    framework, either yours or someone’s else
    ● benefit: simple, small code just for the
    requirements specific to your app
    

    View Slide

20. Example 1:
    

    View Slide

21. Example 2: Meteor
    ● if a user is looking at a list of records,
    obviously when a new record is added, the
    user should see it
    ● & if he changes a record, obviously other
    users looking at it should see the
    modification
    ● so why should you implement this?
    

    View Slide

22. Example 2: Meteor
    ● So what's declarative here?
    ○ You are left with just declarations:
    ■ What data the server publishes
    ■ What data the client subscribes to
    ■ What data is shown in the UI
    ● See this list of 8 open-source apps written in
    Meteor: http://meteorapps.co
    ○ Except for one they are all well written & worth
    learning from..
    

    View Slide

23. Thanks for listening. @dibau_naum_h
    

    View Slide