TODO

Important:
* Precheck boxes do not work in Standalone. The prechecked boxes are only loaded when the data first comes from the EHR, to prevent accidentally overwriting the doctor's choices. Since there is no load out of the EHR, this step never happens. Consider adding a button to Standalone to run this manually.
* Session handling does not work correctly
  - session expires even if "renew session" is clicked
  - session sometimes expires early, seems related to loading/viewing another page, including refpages


Mid Priority:
* Consider moving bin/db-create-portal-tables.sh to the valportaal repo. The tables actually belong to the valportaal, although symlinking them to adfice is handy since adfice needs to write to those tables as well.
* Can we implment a better fake OAuth for testing?

* Getting data out of EHR
  * Look into free FHIR servers and see if we can create a version of the ETL that can be included in the main repo
    https://confluence.hl7.org/display/FHIR/Public+Test+Servers
* Can we test whether the popups appear?
* tidy test data; e.g. a patient with no meas can't happen in Real Life.
* See if we can get rid of <span> headers in nonmed advice. (I think they are not used any more.)
* Can we write any meaningful test for fivepages.js print_page()? It has one job...

* Each acceptance test should start an instance of adfice-webserver
  * refactor adfice-webserver-runner to be more testable
  * refactor the client js (static/patient.js) to be more testable
* automate test reconnect behavior on web-socket drop
  (or adfice-webserver-runner restart)
* specify DB port, separate instance for local human testing, automated testing
* specify DB per-test, or otherwise ensure:
  * tests can be run in parallel
  * data mutations in one test do not impact a different test
  * ensure each test operates on separate data
* CI tests should use HTTPS too, possibly self-signed certs
* the "consult view" button could be renamed "Consult vernieuwen" in order to
  remove the items which have been un-checked/disabled
  * the students suggested using a "reload" icon instead
* record changes between views, connections and disconnects
  * see print logging
* styling, e.g.: viewers in yellow, hidden when 1
* test-cafe : get the innnerHTML out of an element
* sanity check choices of checkboxes
  * e.g.: "stop medication A" && "continue medication A"
* write a log archival script
  raise priority if this looks like it is growing too fast:

      SELECT table_name
           , table_rows
           , ROUND((data_length + index_length)/POWER(1024,2),1) AS size_mb
           , (data_length + index_length) AS size_bytes
        FROM information_schema.tables
       WHERE table_schema='adfice'
    ORDER BY size_bytes DESC;
* make black-and-white printer-friendly css

Low Priority:
* functions created for dataentry2 in message.js have a lot of duplicate code that can probably be refactored
* Rewrite system to work with EHR's proprietary browser - seems to not support web sockets. Or wait 'til 2024 when browser should be upgraded to a 21st-century browser.
* naming consistency thoughout the css
* naming consistency thoughout the javascript
* naming consistency thoughout the html
* tidy css page and make naming more consistant
* testcafe checkboxes contained in display:flex divs are not clickable
  * open a bug with a tiny reproducible example
* the "debounce" of typing input does not work reliably with headless test-cafe
  * adding a "speed:0.1" worked around it, but we should do better
* medications seem to be logged twice on load
* in adfice.js move functions into objects such that if they are called without
  a "this", then a 'not defined' error is thrown right away rather than failing
  at some point in the middle of the function

---------------------------------------------------------
Requested changes (deferred to a later version):
1. System displays different formulations with the same ATC as different medications.
  - Since the advice is the same but the actions might be different, consider combining the display somehow
2. Medications are usually changed gradually over multiple visits. Ideally the system would support multiple visits and changes over time.

---------------------------------------------------------
Requested (not required) changes to rules:
* Rule 50: Split to personalize by age
* Rule 45: Split to personalize by +/- OH.
* Rule 45: Split into using enalapril/something else
* See if there is sufficient evidence to suggest mirabegron as alternative for G04BD
* See if there is sufficient evidence to suggest testerosterone 5 aplha reductase inhibitors (dutasteride) as alternative for G04CA

Important, but unable to do yet:
* Once we know something about how authentication will work, we may need to do
  something to prevent users without authentication from loading the data.
  * session expiration? browser caching? detect back button, return to page?
* Sending data to research database (right now we have some tables in our own database that mimic the research db tables)
* Users wanted a "close this tab" button on each refpage that closes the tab. However, the method that works on IE is considered bad practice, and might not work in the actual deployment environment. We can look into this in the actual deployment enviornment.
---------------------------------------------------------