This software allows for the visualization of disease spreading in hospitals for both bacteria and viruses.
It consists of
- a web server built on Node.js ("backend"), which handles connections to the data sources and performs computations on it, and
- a website ("frontend") served by the same Node.js instance for the actual interaction with the data.
See [Structure of the project](#Structure of the project) for details.
(Planned) features of the system:
- Mobile usability of the frontend
- Build and execution system potentially supporting other languages like C++, Python, R, etc. for algorithmic parts
- Extensible to future data sources, procedures, and - very importantly - more visualization modules
- Useful documentation and quality assurance via continuous integration (CI) with Gitlab
- Most IDs which are used are strings and may not contain any commas, as it uses comma-separated ID lists internally (when communicating with the SQL DB).
- The
IgdSqlDatabaseSource
does not reliably reflect time zones as the underlying data base does not take care of it sufficiently and always runs in local time - This application is not engineered for security in any way. It assumes to be executed in a trusted (virtual) network/environment only.
- Dates being processed must all lie between the years 1980 and 2100. This is an intentional restriction to make sure no bogus dates are being processed, e.g. Unix Epoch 0 or the SAP placeholder year 9999. The restriction is easy to lift if required: Simply change the relevant bounds in the schemas or remove them entirely in
src/server/data_io/type_declarations/*.json
. Searching for"value": "1980-01-01T00:00:00Z"
and"value": "2100-01-01T00:00:00Z"
should do the job. It might be required if anonymization of the sensible medical and health databases introduces such values.
The software is intended to be run on modern Linux systems only. It will probably also run on Windows 10 or other major platforms supporting Node.js & Typescript. Components wich might cause issues include the foreign language libraries. See src/server/foreign_libraries/README.md for details.
The web server requires a recent Node.js version. It is designed and known to be working with version 14 (LTS) and will probably work with future versions too. See here for more information on Node.js versions.
Managing multiple versions is easy with the Node Version Manager (NVM), so we will use it in this manual. You can also install Node.js differently and then skip this section.
# See https://github.com/nvm-sh/nvm#install--update-script on how to install NVM
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.35.3/install.sh | bash
# install version 14
nvm install 14
# in any new shell, do the following to use the correct node version
nvm use 14
# or set it as the default (the first version that is installed will always become the default, so this is for more advanced usage)
nvm alias default 14
The actual installation of the main parts is quite straightforward, although the install
command might take a couple of minutes:
cd to/change/into/a/new/directory
git clone [email protected]:highmed/infectioncontrolsystem2.git
# change into the project
cd infectioncontrolsystem2
npm install
Now everything should be ready to be built and run:
# to build all at once
npm run build
To build all parts individually, please have a look at the definition of the build
command that can be found in package.json
. Alternatively, details are given in [the section on technical details](#Build process).
Important: Afterwards, please follow the instructions in the foreign libraries documentation to set up the parts written in foreign languages.
Before you start the server, please check the configuration to set the desired ports. If not changed, sensible defaults will be set. In any way, the starting server will print instructions on how to visit the graphical user interface (GUI) via the browser in a colorful box to the terminal it is started from. You should be able to click the URL that is shown and should be directed to your standard browser.
After the following command, the webserver should start:
# run the server
npm run start
See [Provided NPM scripts and Compilation](#Provided NPM scripts and Compilation) for more useful commands like running in development mode and exercising unit tests.
The configuration can be found and modified in src/server/config.ts
. The HTTP & HTTPS ports can also be set via environment variables.
This section contains information on why certain decisions were made and how stuff works under the hood.
These commands are defined in the package.json
file. You can easily define new ones, just remember to document them below. They can be used with npm run <COMMAND>
.
There are currently no special commands needed for this.
command | description |
---|---|
build |
This builds both the server and the client, concurrently. |
build-client |
This builds the client (i.e. the website/frontend) using webpack. It does stuff like Javascript transpilation for compatibility and performance (using Babel & Webpack), reducing the file count by combining stuff like all JS or CSS code from different files and obfuscation. |
build-server |
This builds the server (i.e. the backend/webserver code). This does two things: (1) it invokes tsc to build the Typescript code according to tsconfig.json and (2) it invokes build-types . |
build-types |
This command constructs Typescript type annotations from the JSON schemes in src/server/data_io/type_declarations/ to typing information in src/server/data_io/type_declarations/generated/ . It automatically deletes typing files for deleted schemes. This has to be called each time a schema is modified, removed, or added. However, this is already done each time by build-server and thus also by build . |
command | description |
---|---|
dev |
This starts both the client and the server in development mode. |
dev-client |
This runs the frontend in development mode. This makes the browser hot-reload on changes to e.g. style and code and restarts the website seamlessly in-place if required. |
dev-server |
This runs the server in development mode. This makes the webserver (backend) restart if server code changes. |
command | description |
---|---|
start |
This starts the webserver. See also [Starting the Server](#Starting the Server). |
test |
This runs all tests. See also [Starting the Server](#Starting the Server). |
command | description |
---|---|
mirror-release |
This "mirrors" the compiled frontend and backend as source code to the Infection Control System 2 - Release repository. |
There are two major parts of this project: the frontend and the backend.
The frontend is written with very standard web technologies built around HTML and React. Code is written in Javascript/ECMAScript and then transpiled to a version of it that is supported in sufficiently many browsers. This is a very standard approach. Styling is done in SCSS (that is the "SCSS" syntax variant of the SASS styling language) which is then compiled to CSS. It was chosen because it is a strict superset of CSS and thus allows novice users to also use plain CSS to get started. It is, despite offering more flexibility than CSS, easy to use, and easy to compile with webpack.
The backend is written in Typescript, which is a typed "variant" of Javascript. It eventually compiles to just Javascript and enjoys lots of tooling support. It was chosen to provide more (type) safety than Javascript for the many data interfaces. Typescript was chosen above other languages (that might have with better type system or language design) as it still allows us to use all the normal web technologies/libraries around Node.js & NPM. JSON Schema was adopted as a verification technique for the data exchanges as most data is represented as JSON, and it is a technology in widespread use. The platform Node.js was chosen due to its popularity and huge ecosystem of libraries.
The following diagram visualizes the rough data flow in the project:
+-----------------------------------------------+
| |
| Browser Frontend |
| |
+-----+-^----------------------+-^--------------+
npm run start +-+ | | | |
| +-----v-+-----+ +--------------v-+--------------+
| | | | |
+---------------v-+ | Browser GUI | | Data Access |
| | | via HTTP(S) | | via Websockets (via HTTP(S)) |
| start_server.ts | | src/public/ | | src/server/websocket_api.ts |
| | | | | |
+---------------+-+ +-----------^-+ +-^---------+-^-----------------+
| | | | |
| initialize | | v +
+-----------------+-----+ ...
Therefore, the src/
directory is divided into the src/public/
folder with the frontend and the src/server/
folder with the backend.
TODO: @Tom
The following schema gives an overview over the data access via websockets:
+-------------------------------+
| |
| Browser Frontend |
| or in theory other systems |
| |
+--------------+-^--------------+
| |
| | Websockets (via HTTP(S))
| |
+--------------v-+--------------+ MySQL Remote Queries +----------------------+
| <----------------------> SQL DB Data Source |
| Data Access <----------+ +----------------------+
| src/server/websocket_api.ts | |HTTP(S)
| | |(JSON) +----------------------+
+---------^---------------------+ +-----------> MAH REST API |
| +----------------------+
+---------v--------+
| GetDataSources |
+------------------+
If something else than the special GetDataSources
procedure is called, the following steps are performed (error handling aborts immediately):
+-------------------------------+
| |
| Browser Frontend |
| or in theory other systems |
| |
+--------------+-^--------------+
| | Websockets (via HTTP(S))
(1) | | (10)
| |
+--------------v-+---------------+ (2) +---------------------+
| +-----> resolveDataSource |
| Data Access | ++--------------------+
| src/server/websocket_api.ts | (3) |
| <------+
| |
| | +-------------------------------------+
| | | |
| | | Some Concrete Data Source |
| | | |
| | (4) | +-----------------------------+ |
| +---------> Validate Parameters | |
| | | +-----------+-----------------+ |
| | | | (5) |
| | | +-----------v-----------------+ | (6) +---------------------------+
| | | | Transform Arguments +---------> |
| | | +-----------------------------+ | | |
| | | | | Remote Database or API |
| | | +-----------------------------+ | (7) | |
| | | | Transform Response <---------+ |
| | | +-----------+-----------------+ | +---------------------------+
| | | | (8) |
| | (9) | +-----------v-----------------+ |
| <---------+ Validate Response | |
| | | +-----------------------------+ |
| | | |
+--------------------------------+ +-------------------------------------+
This section contains tutorials on how to extend various components of the system.
Validation is performed using JSON Schemas. Information on how to use JSON schemas can be found in the tutorial on the official website. We use ajv for the general validation and the plugin ajv-moment for validation of dates. Please referer to the libraries for documentation.
To add a new schema called TheSchemaName
, two things must be done: (1) the actual schema must be placed in src/server/data_io/type_declarations/TheSchemaName.json
and (2) the schema must be referenced for ease of use. Referencing makes the schema available in the function ensureIsValid()
. The schema name is the ID given within the schema (and not determined by the file name, although please keep them the same for the sake of sanity!).
(1) When creating the JSON file of a new schema, copying an existing one from src/server/data_io/type_declarations/
is probably a good starting point. Make sure to modify the "$id"
attribute (probably right at the start of the schema). Now it's probably good to execute npm run build-types
in order to generate an interface in src/server/data_io/type_declarations/generated/TheSchemaName.d.ts
(there is no need to remember that specific file).
(2) Referencing the schema can be done in src/server/data_io/types.ts
. You can mainly follow the code for the existing schemas for procedure arguments and result data. The detailed steps are:
- Import the actual JSON schema by adding
import * as TheSchemaName_JSON from "./type_declarations/TheSchemaName.json"
to the top of the file. - Add the
TheSchemaName_JSON
to eitherARGUMENT_SCHEMAS
orDATA_SCHEMAS
. Alternatively, you can also create a new group with your new schema and add the new group to theAjv.Options.schemas
below. - Import the interface/typing information by adding
import {TheSchemaName} from "./type_declarations/generated/TheSchemaName"
to the top of the file. - Export the type similar to the other ones. This makes you new type
TheSchemaName
importable from the filesrc/server/data_io/types.ts
, where all other declarations live.
All data sources must extend the abstract class AbstractDataSource
(found in src/server/data_io/abstract_data_provider.ts
) and are usually placed in src/server/data_io/concrete_data_providers/
.
The documentation of AbstractDataSource
describes how the class should roughly behave. It should suffice to implement the abstract data retrieval methods. As part of that, you might have a look at existing data sources.
In the end, make sure to add tests to test/test_all_data_sources.ts
by extending PROCEDURES
.
New procedures can be added by first modifying the AbstractDataSource
in src/server/data_io/abstract_data_provider.ts
. Here, add a method to the end of the class, analogous to the other methods: public abstract async Name_Of_The_Procedure(parameters: Arguments_Whatever): Promise<ResultingDataType>
. Also add an entry to AbstractDataSource::MAPPING
to allow for dynamic calls via AbstractDataSource::callByName
. Finally, the actual procedure must be implemented in all data sources, which may explicitly throw and error if they shall not or cannot support it.
In the end, make sure to add tests to test/test_all_data_sources.ts
by extending PROCEDURES
.
TODO: add link
- The archetypes can be found like this: visit the HiGHmed Clinical Knowledge manager -> select "Use Case Infektionskontrolle" at the top as "project" -> select the orange tab "Templates" on the left -> expand "EHR Templates -> Composition"
- GitLab repository of queries
- Personaldaten-Abfrage -> noch nicht modelliert in CKM, Vorhaben das als "Demographie"-Archetypen zu modellieren, aber eigentlich Trennen der Demographie von Klinischem, Verknüpfung über ID; abwarten aber dann vermutlich möglich, das gibt es auch eine rechtliche Diskussion, an manchen standorten ggf. nie Daten drin, weil sie strikt getrennt sind
- Stronger validation, i.e. sorted-ness, ...
- Linter?
- deploy (started in CI file)
- License & license policy
- Labordaten validierung: wenn befund da => dann KeimID auch da