Skip to content

Latest commit

 

History

History
226 lines (163 loc) · 9.67 KB

README.md

File metadata and controls

226 lines (163 loc) · 9.67 KB

im-tables

CircleCI Clojars Project

A result viewer for Intermine data.

Development goals

  • Simplify im-tables for future developers
  • Build in clojure for easy integration with Intermine's new UI
  • Decrease file size (current im-tables is around 1.2mb uncompressed)
  • See dev approach before attempting to make code changes.

Usage from within ClojureScript

There are two ways to render an im-table. You can mount the component passing a location vector and dispatching the load event yourself.

[im-tables.core/table [:some-location :within-app-db :to-store-values]]

;; In the `:component-did-mount` lifecycle or other suitable place.
(re-frame.core/dispatch [:im-tables/load [:some-location :within-app-db :to-store-values]
                         {:service {:root "beta.flymine.org/beta"
                                    :model some-intermine-model         ; Optional
                                    :summary-fields some-summary-fields ; Optional
                                    }
                          :settings {:pagination {:limit 10}}
                          :response some-response-of-tablerows          ; Optional
                          :query {:from "Gene"
                                  :select ["symbol"
                                           "secondaryIdentifier"
                                           "primaryIdentifier"
                                           "organism.name"
                                           "dataSets.name"]
                                  :where [{:path "Gene.symbol"
                                           :op "LIKE"
                                           :value "M*"}]}}])

The other way is to pass in the entire map to have the im-table dispatch the load event automatically on mount.

[im-tables.core/table {:location [:some-location :within-app-db :to-store-values]
                       :service {:root "beta.flymine.org/beta"
                                 :model some-intermine-model         ; Optional
                                 :summary-fields some-summary-fields ; Optional
                                 }
                       :settings {:pagination {:limit 10}}
                       :response some-response-of-tablerows          ; Optional
                       :query {:from "Gene"
                               :select ["symbol"
                                        "secondaryIdentifier"
                                        "primaryIdentifier"
                                        "organism.name"
                                        "dataSets.name"]
                               :where [{:path "Gene.symbol"
                                        :op "LIKE"
                                        :value "M*"}]}}]

Some of the values are optional, and if supplied they will not be fetched when the component is mounted. This allows you to fetch a resource once and share it across components, or to render a table of query results that have already been fetched.

Development

There are just a few core concepts that must be understood before developing on the project. Seriously. Keeping these in mind will save you money spent on headache medication.

1. im-tables is not just a React component, it's also an application., and :location is everything

It uses re-frame as an in-memory database to store its values, and integrating a re-frame application into a parent re-frame application can be tricky. Why? If a parent application (BlueGenes) mounts two table components, events fired from one table to store values in app-db will override the other's state.

This makes the :location key of the options map crucial. It is a vector of your choosing, and it tells any given table (or tables!) to only update values in that sublocation of app-db. The :location value should be passed to every single view in im-tables, and then passed to every single event handler as its first argument. Same goes for subscriptions. This is a manual process and no checks can be made to make this happen. If an event-handler using the sandbox interceptor is not passed a location vector as its first argument then you will likely not see your updates because they're being made to the nil key in app-db.

The sandbox interceptor extracts the sublocation of app-db and provides it to your event handler as if it were at the top level. This allows you to write your event handler functions as if im-tables was a standalone application rather than repeatedly reaching into a specific location in app-db. For example:

Using the sandbox interceptor, the following event can be written as such:

(reg-event-db
  :show-overlay
  (sandbox)
  (fn [db [_ location value]]
    (assoc-in db [:cache :overlay?] value)))

Whereas a version without the sandbox interceptor is more verbose:

(reg-event-db
  :show-overlay
  (sandbox)
  (fn [db [_ location value]]
    (update-in db location assoc-in [:cache :overlay?] value)))

That's not a huge for simple handler functions, but it pays out when you have more complex ones that use get-in and update-in to access values in many parts of your app-db, and the function can be tested without having to prepare a mock app-db with nested values.

Subscriptions don't have middleware so they need to take the table's app-db location into account:

(defn glue [path remainder-vec]
  (reduce conj (or path []) remainder-vec))

(reg-sub
  :main/query-response
  (fn [db [_ location]]
    (get-in db (glue location [:response]))))

2. The location is used by most subscriptions and is often bound in a component's outermost closure.

Another opportunity for a headache! Form 2 components are functions that return functions. When the component updates, the returned function is called to update the view. Values bound to the outermost function will never return anything but their original value. In this example cake will never change.

; The component WILL NOT update when cake changes
(defn [cake]
  (fn []
    [:h1 cake]))

; ...whereas this component will
(defn []
  (fn [cake]
    [:h1 cake]))

You will often see location passed to the outermost function, but don't be caught in the trap of binding more volatile values there as well.

3. The view is optimised to render quickly

When a table first loads it only renders the first page of the query results as bog standard html with minimal components. This allows many tables to be rendered to, say, a report page while reducing load on the browser. Mousing over the table triggers a change that forces the table into "React" mode when the table cells become more complex.

Building

Initial setup

To compile the css and run the unit tests, you'll need Node.js and npm. Once these have been installed, run the following.

npm install

Quickstart

These commands are explained in more depth below, but if you know what you want here's a quick reference of the most useful ones.

lein dev         # start dev server with hot-reloading
lein repl        # start dev server with hot-reloading and nrepl (no clean or css)
lein deploy      # build prod release and deploy to clojars

lein format      # run cljfmt to fix code indentation
lein kaocha      # run unit tests

Compile css:

We use less to write our styles.

You can compile the Less sources to CSS with the command below:

lein less

It will run in watch mode, recompiling the CSS after every change until you exit it with Ctrl+C. The browser page will automatically reload the updated CSS.

When using the other lein commands to build or deploy, the less sources will automatically be compiled for production mode.

Run application:

lein clean
lein figwheel dev

Figwheel will automatically push cljs changes to the browser.

Wait a bit, then browse to http://localhost:3448.

Run tests:

Make sure to first run npm install to install the dependencies required to run the unit tests. You can also add --watch to the command below to automatically re-run tests when saving files.

lein kaocha

Note: Running tests requires you to have a Biotestmine running locally. The easiest way to achieve this is by using intermine_boot.

Production Build

Deploying to Heroku

lein clean
lein uberjar

That should compile the clojurescript code first, and then create the standalone jar.

When you run the jar you can set the port the ring server will use by setting the environment variable PORT. If it's not set, it will run on port 3000 by default.

To deploy to heroku, first create your app:

heroku create

Then deploy the application:

git push heroku master

To compile clojurescript to javascript:

lein clean
lein cljsbuild once min

Releasing a new version

The release process is a combination of the above commands, with some additional steps. Generally, you'll want to do the following.

  1. Update the version number in project.clj.
  2. Prepend changes since previous version to CHANGELOG.md.
  3. Commit this change and tag it using git tag -a v1.0.0 -m "Release v1.0.0", replacing 1.0.0 with your version number.
  4. Push your commit and tag using git push origin followed by git push origin v1.0.0 (again replace 1.0.0 with your version number). Make sure that you push to the intermine repository, not just your fork!
  5. Deploy a new uberjar to Clojars with lein deploy.