CHANGELOG.md

Ongoing

geoserver cache: https://octopus.com/blog/defining-tomcat-context-paths

Long term work

• Remove plotly completely, use d3 only. Requires that createBarChart and createGroupedBarchart are not called from detailPopupHTML. Reason: d3 cannot handle dom-nodes that don't exist yet. For the new d3-graphs we need a dynamicPopupComponent. And for a dynamicPopupComponent we need to return a ProductCustomLayer, not a VectorLayerProduct.
• Integrate backend for content-redaction
• Old frontend: modularize
• Cache: move from file-cache to geoserver-cache
• Backend: known race-condition
  • backend is being sent some state for an execute-request
  • that state contains a reference to data that has by now expired
  • so the backend can't resolve that data
  • Solution:
    • backtrack: from the data-reference, find out what state was used to create the data, and reproduce it
    • do this recursively
    • requires database with metadata about the references
• Compare frontend:
  • Problem: user needs to define order of composites the same in wizard-converter and in map-converter.
  • Otherwise the top layer on map is not the top layer in layercontrol
  • That's because map-converters and wizard-converters are completely independent.

[2.0.7]

Features

• merged in new frontend.
• icons for different modes
• showing version number
• showing mode upon selection
• activate right side once left all completed
• showing currently selected eq in config-bar
• eq labels formatted
• users can now close popups
• popups contain helpful info
• ? icon for each mode
• map-component: layers only re-built if they have actually changed
• eqLegend: closer together
• popups:
  • correct popup shown when multiple layers at same time
• no longer jump to location
• for some steps, only one layer visible at the same time
• update eq style upon dropdown select
• translate html tables
• parseAPIScenariosIntoNewState
• tsunami damage
• tsunami sim
• sysrel

Bugs

• Fixed race condition in eq- and ts-damage popups

2.0.6 (Jun 11 2023) Compare prototype working

Features

• Overall:
  • all components of architecture now dockerized
• Backend
  • all ScenarioFactories can now register async conditions which must be fulfilled before the scenario may be used.
  • Execute: failing fast if data is outdated.
• Compare frontend
  • link click on eqSelection with form
  • mirrorClick rule now allows include and exclude of compositeIds
  • styles feature differently on click
  • Custom converter for wizard/eqSim
  • exposure now displayed
  • exposure wizard now configurable
  • exposure legend and styling
  • errors displayed in ui
  • on error: retry
  • simplified error-handling: backend-client retuns a Maybe, doesn't throw.
  • update autopilot:
    • needs not use that "first-step" condition anymore
    • can define on per-step-basis if autopilot required
  • auto-piloted steps are no longer interactive
  • layer control, syncing visibility between wizard and map
  • Icons
  • translation service
  • Map component: manually triggering change detection after adding dynamic component
  • Exposure barchart: fixed hover effect
  • Wizard: progress bar
  • Fix: Popups in compare-two-scenarios: if one side cannot create a popup, a closePopup action is no longer dispatched to global
  • eq dmg econ popup
  • eq sim popup and legend

Bug fixes

• Compare frontend
  • Map zooms in to first available step right after selecting a scenario

2.0.4 (May 16 2023) Created monitor and new frontend

Features

• monitor:
  • created service that regularly executes all riesgos-services
• backend:
  • now provides optional default values for steps' inputs.
  • tickets now stay alive for a certain time after they've been fetched
  • added execute?skipCache=true option for execute-requests
  • added runall.ts which runs all steps once every hour.
  • added expires headers
• frontend:
  • now allows multiple (dynamic) legends for one layer at the same time. Applied to eq-catalogue.
  • now has circle-legend
• Compare frontend
  • all code now in modules
  • map allows per-step data-converters which are semi-automatically injected
  • now displays available eq's, too.
  • using rules to decide what exactly to do in reducers
  • user can chose between rule-sets
  • map has click-behavior
    • add popup to map on click
    • can now close popups
    • popups get features at click location as input
    • composites can do arbitrary things on click
  • Rules now calculated from rule-set-name by dedicated service
  • LayerComposite methods are non-anonymous: allows accessing all properties with this.
  • Custom style for eq-selection layer.
  • Wizard now converts from dynamically selected converter, too, just like map does.
  • More than one step can be focused
  • Arbitrary screen-partitions possible
  • Wizards now have a default converter to fall back to
  • error-handling during process-execution
  • click on step toggles focus
  • hides layers of un-toggled steps
  • parameter-selection shows labels in dropdown
  • insideOne ruleset: data now mirrored
  • auto-pilot only started when rules allow it
  • legends now displayed - if present
  • popup: increased close button clickable area
  • dropdown now selects value when clicked on map
  • wizard expanded from beginning: did need to move share up in filter-obs
  • reducers: parseAPIScenariosIntoNewState: strict typing
  • openlayers now runs outside of angular-zone

2.0.3 (Mar. 9 2023) Bug fixes

Bug fixes

• layer- and wizard-services no longer cause multiple ui-updates on every state-change (reason: base-observable now shared)
• click on map only fires click-handler once.
• removed superfluous console-logs
• literal parameters of eq-catalog were not accepted
  • because no userDataProvided action triggered when user edits literal values
• logging threw error on circular objects.
• race condition in making post-eq-damage mapable.
• wrong style and legend for chile-damage.
• should have fixed NS_BINDING_ABORTED (happend in Firefox, presumably because too many posts at the same time).
• error in one process does not stop processing of other processes.

2.0.2 (Feb. 2 2023) Redactional

Features

• Discrete legends: slight outline around each field
• Continuous legends: added marker-line
• Nicer styling for layer-details: no word-breaking, less distance between paragraph and legend.
• Styling:
  • Fixed positioning eq-selection
  • Legend titles smaller
  • Less spacing around form controls, paragraphs
  • Small tables a little bigger
  • Simplified EQ ID display
• Fixed some dead links, added some data-sources
• Reconfigure button disabled for a second so users can't click it accidentally when the step-menu pops open.
• EQ-Catalog: added the parameters mmin, mmax, zmin, zmax, and p again.
• EQ-Selection: showing id and magnitude for each element in dropdown.
• Exposure popup: now highlights current building type and gives plain-text version of building class name.
• Added new and corrected old data-sources .

Bug Fixes

• Ecuador: fixed path-error for Lahars with P_ex < 10%
• Backend sometimes doesn't stop an erroneous process
  • Might be because it attempts to send an email, which did throw an error itself (separate bug, now resolved). So has to send another email ... which keeps repeating.
  • May be re-opened if the error occurs again.
• Frontend did sometimes not display error messages from backend. Reason was that caught observable could not be completed to get scenario-id and step-id. Fixed by adding that information to the thrown error.
• Damage-popups did not show no-data.

2.0.1 (Jan. 4 2023) Bugfix

Features

Bug Fixes

• Bugfix: bad use of d3-color in damage-popup.
• Bugfix: when going back to eq-sim-service from further upstream, its products are not being highlighted. Reason: mistakenly gave the layer productId = product.description.id, should have been productId = product.id

Breaking Changes

2.0.0 (Dec. 8 2022) Stable version of new frontend and backend

Features

• Added testing-suite to frontend's ngrx-state-manager

Bug Fixes

• Legend ecuador/damage post-lahar: fixed cut-off
• Frontend-state: fixed race-conditions in calculating state when two processes run in parallel

Breaking Changes

2.0.0-alpha (Dec. 1 2022) All orchestration-logic in backend

Large refactor.

Motivation was a problem seen when attempting to use Riesgos at other institutions.

• Replacing services was hard. Reason: new services provide data in new formats, and often downstream processes require a very specific input format.
• Solution: we provide a layer between the frontend and the webservices. This layer - a node.js-backend - serves to handle all the harmonization between the actual webservices.
• Any required duct-taping of service-outputs can now be handled in this intermediate layer.

Previously there was already a lot of harmonization-code in the frontend's WpsProcess classes. But that meant that the frontend would manage webservices, their orchestration, their harmonization, in addition to all its usual tasks (state-management, error-displaying, converting output-data into a format that can be displayed on the map, ...)

Features

• Moved step-logic into backend

  • Justification:
    • Frontend was over-full, hard to maintain
    • Backend can now run steps as CI
    • Central log of events and errors
    • Most important: Easier harmonization of webservices.

• No longer based on wps'es, but on abstract steps

  • Many steps do not involve calling a webservices, and many more involve calling several.
  • Setting up a webservice is hard, writing some js code in backend is relavitely easy.
  • Most important: often steps had cross-dependencies. Having them all in the backend allows the developer one point where changes to the harmonization can be made.

• Frontend: ngrx-state is simple and abstract

  • Individual components may add additional information to state using Augmentors.
  • But generally, information that is only relevant to one component should not pollute the global state.

• Backend: steps may not run side-effects

  • Based on problem: Selecting an eq from the catalog-output
  • Decision: The backend will not set the options parameter of the userChoice parameter.
  • Justification:
    • The catalog's output does already constitute the possible options for userChoice.
    • The alternative means that we'd allow side effects in the backend.
    • This allows for so many complications that it's not worth the - already questionable - feature.
    • Examples of potential complications:
      • side-effects that add new values must be cached, side-effects that only add new options must not
      • if side-effects may output options to a datum, it's no longer guaranteed that new data consists 100% of resolved data

• Data passed per reference per default

  • Riesgos uses a lot of data
  • All that data clogs up browser memory, makes ngrx-state too big to be easily re-created, and takes long to download
  • So now we pass data per reference. Those references are resolved lazily. Once resolved data is also cached.