no-stack
v0.0.43
Published
Set of React utilities and helpers for no-stack
Downloads
1
Readme
no-stack
NOTE: this is being released in ALPHA. Not yet stable!!!! We are rebuilding as TS and with a lot more functionality.
no-stack
lets you build a full stack application using nothing but static front
end code. You can even run it without a server, for instance using an AWS S3 bucket
enabled as an endpoint.
To get the full value of no-stack
, you need to register a platform. Currently,
you must do that at www.nostack.net.
Then, you can build your virtual back end in one of two ways. (1) You can use our api (api.matchlynx.com). (2) You can run your app as a user with "moderator" status.
You start by declaring types of users called user classes. Then, you give each class
actions. You can build whatever pages you like. You can declare and use data Types
which have Instances
. Types
also have Assns
(associations) between each other.
A page can use data via Sources
. Each source has some minimal necessary Assns
and the Types
that you need. You can also create Constraints
on the Instances
returned.
This repository contains the no-stack
helper components and functions. Documentation
coming shortly...
Getting Started
Prerequisites
Be sure you have the latest stable version of Node & NPM installed. Depending on your machine, you would probably need to fix permissions for NPM.
You might also need to install TypeScript globally.
Install Dependencies
no-stack
requires React & Apollo as dependencies.
To install React, use create-react-app
and bootstrap your app by following
these directions.
Then cd
into your app folder and install Apollo:
npm install apollo-client graphql react-apollo apollo-link \
apollo-cache-inmemory --save
Install the rest of no-stack
's dependencies:
npm install apollo-fetch apollo-link-context apollo-link-http axios formik \
graphql-tag jsonwebtoken react-graph-vis react-spinkit styled-components \
uuid --save
Install no-stack
via Github
no-stack
is not yet available on NPM so you would need to install it directly from
Github via SSH.
Add dependencies entry in the app's package.json
file:
"dependencies": {
...
"no-stack": "git+https://[email protected]/NoStackApp/no-stack.git"
}
The on the command line, run:
npm install
Installing a Specific Version
You may opt to install a particular version of no-stack
by adding the version
tag at the end of its dependencies entry in package.json
:
"dependencies": {
...
"no-stack": "git+https://[email protected]/NoStackApp/no-stack.git#v0.0.12"
}
Initial Set-Up
This guide will create a basic set-up for your React & Apollo instances. If you need a more complicated set-up, please refer to their respective official documentation.
Setting up React and Apollo
Edit src/index.js
file like so:
import React from 'react';
import ReactDOM from 'react-dom';
import { ApolloProvider } from 'react-apollo';
import './index.css';
import client from './client'; // we will create this file
import App from './App';
import * as serviceWorker from './serviceWorker';
// wrap your app inside ApolloProvider
ReactDOM.render(
<ApolloProvider client={client}>
<App />
</ApolloProvider>,
document.getElementById('root'),
);
// If you want your app to work offline and load faster, you can change
// unregister() to register() below. Note this comes with some pitfalls.
// Learn more about service workers: http://bit.ly/CRA-PWA
serviceWorker.unregister();
Then edit src/client/index.js
:
import { ApolloClient } from 'apollo-client';
import { ApolloLink } from 'apollo-link';
import { InMemoryCache } from 'apollo-cache-inmemory';
import { createAuthLink, httpLink } from 'no-stack';
const PLATFORM_ID = 'your-platform-id';
const authLink = createAuthLink(PLATFORM_ID);
const link = ApolloLink.from([authLink, httpLink]);
export default new ApolloClient({
link,
cache: new InMemoryCache({
dataIdFromObject: object => object.id,
}),
});
Cache Set-Up
To allow no-stack
's helper functions (discussed below) to access and update
Apollo's cache, dataIdFromObject
must be set up like we did above, which we
repeat here:
const client = new ApolloClient({
// ...other options...
cache: new InMemoryCache({
// ... other options ...
dataIdFromObject: object => object.id,
}),
});
Bootstrapping Your Application
To enable no-stack
's features in your application, you need to wrap it
inside the <NoStackProvider>
component. You may do so by adding
it in our previous src/index.js
file like so:
import React from 'react';
import ReactDOM from 'react-dom';
import { ApolloProvider } from 'react-apollo';
import { NoStackProvider } from 'no-stack';
import './index.css';
import client from './client'; // we will create this file
import App from './App';
import * as serviceWorker from './serviceWorker';
const PLATFORM_ID = 'your-platform-id';
// wrap your app inside NoStackProvider
// then wrap NoStackProvider inside ApolloProvider
ReactDOM.render(
<ApolloProvider client={client}>
<NoStackProvider client={client} platformId={PLATFORM_ID}>
<App />
</NoStackProvider>
</ApolloProvider>,
document.getElementById('root'),
);
// If you want your app to work offline and load faster, you can change
// unregister() to register() below. Note this comes with some pitfalls.
// Learn more about service workers: http://bit.ly/CRA-PWA
serviceWorker.unregister();
The only required props of <NoStackProvider>
is your platformId
and the client
.
Usage
Authentication
no-stack
provides a helper components, <LoginForm>
& <LogoutButton>
to automate
login & logout. Sample usage:
import React from 'react';
import { LoginForm, LogoutButton } from 'no-stack';
const SomeComponent = () => (
<NoStackConsumer>
{({ currentUser, loading }) => {
if (loading) return 'Loading...';
if (!currentUser) {
return <LoginForm />;
}
return <LogoutButton />;
}
</NoStackConsumer>
);
export default SomeComponent;
Note: <LogoutButton>
hides itself automatically if user is not logged in.
<NoStackConsumer>
Component
To access the authentication state of the user, you will need <NoStackConsumer>
.
Sample usage was illustrated above.
It also provides some helper methods if you want to customize your login/logout behavior.
<NoStackConsumer>
Function as a Child (FAAC) Props
platformId
- The current platform's IDcurrentUser
- The current user object{id: 'some-id', name: 'some-name', role: 'some-role'}
loading
- Boolean.true
if currently validating user with API.login({ username: 'some-username', password: 'some-password' })
- Helper function to log in a user with credentials. If successful, it setscurrentUser
and stores auth tokens in browser'slocalStorage
.logout(callback)
- Helper function to log out user. RemovescurrentUser
and all auth tokens. Executes provided callback function after logout.
Alternative: withNoStack
Higher Order Component
You may interchage <NoStackConsumer>
with withNoStack
HOC and achieve the
same effect:
import { withNoStack } from 'no-stack';
const SomeComponent = ({ currentUser, loading, login, logout }) => {
if (loading) return 'Loading...';
if (!currentUser) {
return <LoginForm onSubmit={login} />;
}
return <LogoutButton onClick={logout} />;
};
export default withNoStack(SomeComponent);
withNoStack
also provides the same props that <NoStackConsumer>
's FAAC provides.
Feel free to use either, whatever suits your preference or situation.
Platform Management Console
After setting up authentication, the second thing you might want to implement
is the <Console>
.
The <Console>
component is provided for the sole purpose of managing your
current platform (i.e. adding, editing, and deleting user classes). When a
moderator is logged in, It displays a transparent button on the lower left
(or lower right) corner of the page you insert it into. When that button
is clicked, a modal showing your dashboard will pop up.
You may place it anywhere on your app, as long as it's a child or further
down the hierarchy of a <NoStackProvider>
. We do suggest that you place it
near the root of your application so that the button is accessible in all your pages.
Sample usage:
import { Console, NoStackProvider } from 'no-stack';
import RootComponent from './RootComponent';
export default const App = () => (
<NoStackProvider>
<Console position="left" />
<RootComponent />
</NoStackProvider>
);
position
prop - Displays your console either on the bottom left
or
bottom right
of the current page (default: left
)
Data Retrieval via the <Source>
Component (Experimental)
no-stack
provides a <Source>
component for retrieving data.
no-stack
uses GraphQL. See this Intro to GraphQL
for more info about using GraphQL, and
GraphQL Queries in particular.
The no-stack
Source
Component is very close to the
Apollo Client Query Component. Both take the
same prop of a query
.
Following is an example usage.
import React from 'react';
import { Source } from 'no-stack';
import gql from 'graphql-tag';
const SOURCE_QUERY = gql`
query SOURCE(
$id: ID!
$typeRelationships: String!
$parameters: String
$unrestricted: Boolean!
) {
sourceData(
sourceId: $id
typeRelationships: $typeRelationships
parameters: $parameters
unrestricted: $unrestricted
) {
instance {
id
value
type
}
children {
instance {
id
value
type
}
children {
instance {
id
value
type
}
}
}
}
}
`;
const sourceId = 'your-source-id';
const typeRelationships = {
'first-type-id': {
'second-type-id': {
'third-type-id': null,
},
},
};
const unrestricted = true;
const YourComponent = ({ someParam }) => {
const parameters = {
someParam,
};
return (
<Source
id={sourceId}
typeRelationships={typeRelationships}
query={SOURCE_QUERY}
unrestricted={unrestricted}
parameters={parameters}
>
{({ loading, error, data }) => {
if (loading) {
return 'Loading...';
}
if (error) return `Error: ${error.graphQLErrors}`;
console.log(data);
return <div>Test Source</div>;
}}
</Source>
);
};
export default YourComponent;
<Source>
Props
id
- the source IDtypeRelationships
- the desired return type hierarchy (see below)query
- the GraphQL tagged query string or file (see below)parameters
- an object consisting of key-value pairs representing the required constraints of the sourceunrestricted
- boolean;true
if data is available to public,false
if otherwise. (default: false)
typeRelationships
Prop
A source can be thought of as a virtual table, like a view in a relational database.
Each type is like a column. In relational terms, the typeRelationships
prop is
like a "projection" (letting you specify which of the "columns" (types) you want
returned in your request.
But, unlike with a table, the data returned here is a hierarchy, where a child type's instances are grouped based on the parent type's instances. So when you specify what you want, you must start with the highest level of that hierarchy (the root), and then list it's children (if any). You then do the same recursively for each subtree.
Therefore, the typeRelationships
is tree, specified as a JSON. The format is as
follows:
{
<parentType>: {
<child1>: {
<child1a>: <child1a-subtree>,
<child1b>: <child1b-subtree>,
...
},
<child2>: {
<child2a>: <child2a-subtree>,
<child2b>: <child2b-subtree>,
...
},
...
},
}
The simplest case would be a single type with no children:
{ doctor_type_id: {} }
. In this case, doctor has no child.
A bit more complex would be a type with one child:
{ doctor_type_id: { nurse_type_id: {}, patient_type_id: {} }}
. In this case,
each doctor will be returned with children including nurses for each doctor and patients
for each doctor.
A third possibility is this:
{
"doctor_type_id": {
"nurse_type_id": {
"patient_type_id": {}
}
}
}
Note that here each doctor will have nurses, and each nurse will have patients.
Usually, a typeRelationships
won't be large at all. (That's probably better style
anyway. In general, no-stack
encourages minimizing state on the front end.)
Some important notes:
If a type does not have child types, represent it by setting the type to
{}
.Neither the choice of fields nor their place in the hierarchy will constrain the data returned. The data for a source at any given moment is fixed, and the type hierarchy simply selects what columns are returned and the grouping. For example, say that a source contains a set of doctors, nurses, and patients. Then all of the doctors contained in that source will be returned whether the type hierarchy has just doctors or doctors and nurses. The reason is that if a doctor had no nurses than he or she would not be included in the source in the first place.
That said, the data returned is distinct within the selected sets. So a small
typeRelationships
might well return fewer instances than a larger one. The rule is that each instance of the root type appears only once, and each instance of a child type appears only once for its parent instance. For instance, if Doctor A has ten nurses, then listing by doctor as the root would show Doctor A once with all ten of his nurses, whereas listing by nurse first would show the same doctor ten times, once for each nurse.
query
Prop
As stated above, The Source
Component is very close to the
Apollo Client Query Component. Both take the
same prop of a query
.
The query prop is the actual query to use. In the case above, it is a gql
specification.
In addition, you can provide a file that includes the complete query specification.
The plan in the near future is to generate such files for you, which you can then insert into your project. In the meantime, you must generate it yourself.
Fundamentally, the query sourceData
is called with four variables (which is provided
by Source
's props discussed above):
$id: ID!
$typeRelationships: String!
$parameters: String
$unrestricted: Boolean!
Fields requested
A call to a GraphQL API allows you as the developer to decide which specific "fields" you want to request to be part of the returned data.
no-stack
uses GraphQL. See GraphQL Query Fields
for an introduction.
Specifically, information about the instances for types in your typeHiearchy
are
returned. For each returned instance you can choose from the following fields in
your SourceData
request:
{
id,
type,
instance {
id,
value,
order
},
children {
...
}
The children
field can recursively contain any of these same fields as deeply as
you desire up to the depth of its typeRelationships
.
The meaning of the fields is as follows:
id
- current type's idtype
- the instance typeinstance
- the instance info. Specifically includes theid
for the instance, as well as it'svalue
and anorder
if one exists. Every instance has an id; some instances don't have values or orders.children
- any child instances for this instance. See thetypeRelationships
explanation above to understand how the relationships are specified there.
Here is an example of the fields requested. This json specifies all of the data to a depth of 3:
{
id
type
instance {
id
value
order
}
children {
id
type
instance {
id
value
order
}
children {
id
type
instance {
id
value
order
}
}
}
}
In general, you should request up to the depth of your typeRelationships
(see above).
Anything less will be missing data, and anything more will not return anything.
<Source>
Function as a Child (FAAC) Props
loading
- boolean;true
if data is still being fetched,false
if otherwise.error
- error object;null
if there's operation succeeds.data
- the data object.queryVariables
- object containing the props provided to the<Source>
(refer to discussion above).refetchQueries
- a single element array containing the object to be fed to Apollo mutation's ownrefetchQueries
property (see mutation discussion below).updateSourceAfterCreateAction(instance)
- a higher-order function function for updating Apollo cache after creating an instance (see mutation discussion below).updateSourceAfterUpdateAction(instanceId, fragment)
- a higher-order function for updating Apollo cache after updating an instance (see mutation discussion below).
<Source>
Inspector
In addition to data retrieval, the <Source>
component also has a data inspector
modal that is accessible to any logged-in platform moderator.
You don't have to do anything special in your code to enable this feature. Once you're
logged in as a moderator via <NoStackConsumer>
, a button with the plus (+) sign
will show on every instance of the <Source>
component on the current page.
The inspector currently only shows the relationships between types of data (nodes) via a type tree graph visualization. But you will eventually be able to manage your data type tree (i.e. add/remove types) using the inspector.
Manipulating Data via Action Mutations (Experimental)
Currently, data manipulation uses Apollo's GraphQL components. no-stack
provides
a mutation helper you can plug into Apollo's <Mutation>
component called EXECUTE_ACTION
:
import { Mutation } from 'react-apollo';
import { EXECUTE_ACTION } from 'no-stack';
const YourComponent = () => (
<Mutation mutation={EXECUTE_ACTION} >
{(executeAction, { data }) =>
if (data) {
console.log(data);
}
return (
<div>
<button onClick={() => executeAction({
variables: {
actionId: 'dummy-id',
executionParameters: JSON.stringify({ someParameter: 'value' }),
unrestricted: false,
},
})}>
Execute Action
</button>
</div>
);
}
</Mutation>
);
Or, via Apollo's graphql
HOC:
import { graphql } from 'react-apollo';
import { EXECUTE_ACTION } from 'no-stack';
const YourComponent = ({ executeAction }) => (
<div>
<button onClick={async () => {
const { data } = await executeAction({
variables: {
actionId: 'dummy-id',
executionParameters: JSON.stringify({ someParameter: 'value' }),
unrestricted: false,
},
});
console.log(data);
}>
Execute Action
</button>
</div>
);
export default graphql(
EXECUTE_ACTION,
{
name: 'executeAction',
},
)(YourComponent);
Depending on its type, the action called with EXECUTE_ACTION
can perform different
mutations, including but not limited to creating/updating/deleting instances, logging
in/out, and user registration.
EXECUTE_ACTION SIGNATURE
| Variable | Description | Value | Required |
| ------------------- | ----------------------------------------------------------------------- | ---------------------------------------------------------- | --------------------- |
| actionId | An existing action ID | ID | Yes |
| executionParameters | An object describing the optional and required parameters of the action | Valid JSON String (use JSON.stringify()
) | Depends on the action |
| unrestricted | Signifies if the action authorization | true
if restricted to logged in users (default: false
) | No |
Updating Apollo Cache After Action Mutations
Apollo implements a front end store for all GraphQL queries that get executed which they call the "cache".
Usually, after calling GraphQL mutations, or particularly in no-stack
's case, action
mutations, the data on the backend and on the Apollo Cache become out of sync. To
keep Apollo Cache in sync, you need to access it directly.
To help make this process easier, no-stack's <Source>
component provides a few
helpers via its render props. (Note: Sample usage code is from the demo Todo App.)
refetchQueries
refetchQueries
is an array to be used with
Apollo mutation's refetchQueries
props.
This is the most convenient way of updating the Apollo cache. But it is also the
slowest, since it requires making a roundtrip HTTP transaction with the no-stack
API.
For usage, see <CreateIsCompletedForm>
.
updateSourceAfterCreateAction(instance)
updateSourceAfterCreateAction
is a higher-order function that returns a function
that fits the signature for the callback used by Apollo mutation's update
property.
As the function name suggests, this is mostly suitable after an action that creates
an instance.
For sample usage, see <CreateProjectForm>
.
Note: <CreateProjectForm>
's parent passes it updateSourceAfterCreateAction
as
the onAdd
prop.
A slightly more complicated usage is found in <CreateTodoForm>
.
updateSourceAfterUpdateAction(instanceId, fragment)
updateSourceAfterUpdateAction
is similar to updateSourceAfterCreateAction
, but
it is meant to be used for updating the cache after an action that updates an instance.
For sample usage, see <Project>
.
updateSourceAfterDeleteAction(instanceId, fragment)
(To be implemented)