mstn / graphql Goto Github PK

Some ideas to extend GraphQL.

• Recursion. As far I know, it is forbidden in current spec because it could yield infinite results. However, it is useful. A recursive query represents a potentially infinite result, but we do not have to send infinite data to the client. Some examples:
  • pagination (e.g. infinite scrolling)
  • cyclic data structures (e.g. I am a friend of my friends). Currently, data are unfolded and cyclic relations are potentially infinite, but folded data have a finite representation as graphs!
• Restructuring. Alias are a simple mechanism to rename query properties. It could have sense also to explore more complex post-processing.

CLI version: 2.3.2
GitBook version: 3.2.3

Error: ENOENT: no such file or directory, open 'undefined/literature.bib'

Probably, it is due to plugin bibtex-citation

Note that if we downgrade gitbook, it works.

gitbook serve --gitbook=2.3.2

http://stackoverflow.com/questions/16212364/how-do-i-display-a-proof-tree-with-html-css-and-or-javascript

• Cardelli-Ghelli
• Molyneux-Sazanov, in particular their "execution semantics"
• Sparql

As a reader I want to find information about content license easily.

Currently, we defined a sort of big step semantics for query execution. It seems convenient to follow a small-step approach.

• Easier to formalise the concept of stuck execution
• Different queries could in theory be executed in parallel and return partial results to the client
• "Lazy loading", i.e. deferring and streaming.

EDIT - added "lazy loading"
New upcoming GraphQL features (e.g. @defer, @stream) allow a sort of lazy loading. Reference

As a reader I want to find a description of the core language with its semantics so that I can use it later as a reference for more advanced topics.

• tree data structure
• types
• queries
• examples

As a writer I want to a simple way to publish and edit content so that I can ship as fast as possible.

• create npm parent project
• create gitbook project
• add katex plugin to gb project
• add reference plugin to gb project
• create gitbook project
• create bash script to build and publish to gh-pages
• add licenses, disclaimer etc

List of GraphQL features

• Queries #4 #12
• Type system #4 #12
• Execution semantics #4 #12
• Inline Fragments #12
• Unions #12
• Fragments
• Alias
• Enums
• Directives
• Introspection
• Interfaces
• Error Handling
• Non Null types

Our assumptions/simplifications

• root type can be any type #12
• Named queries with parameters -> arguments are free vars in query #12
• mutation -> mutation = query with side effects, for now, we do not have any special syntax to distinguish them #12
• no special type for input objects #12

Desirable properties (or what can we say about the abstract lang?)

• safety
  • progress
  • preservation
• uniqueness of typing

EDIT changed to epic, not so easy

First, I would like to verify properties of the abstract syntax in a formal way. In this way I can double check the pen-and-pencil results of the notes. Writing derivation trees in html and css is otherwise error prone.

Then, I would like to test the formal semantics on a test suite. In other words I would like to be sure that the abstract language I am defining is as close as possible to (one of) the actual implementation(s) of GraphQL. The easiest way is to extract an interpreter from a Coq spec and try it with existing test suites.

A complex demo of meteor+graphql with real world features

• pagination/infinite scrolling
  • smart scrolling, remember last position
• advanced search on a list of items
• counters grouped by categories
• perm links

As a reader I want to have all cited articles in a single page.

See also:

• bibtex plugin for gitbook