shakemap-modules 1.2.2__tar.gz → 1.2.3__tar.gz

{shakemap_modules-1.2.2 → shakemap_modules-1.2.3}/CHANGELOG.md

@@ -1,10 +1,131 @@
 # main
 
+## 1.2.3 / 2026-09-06
+- Modules refactor
+    - Refactored and reviewed ShakeMap top-level modules for consistent
+      standalone/shake behavior, resource handling, and CoreModule
+      argument parsing.
+    - Fixed standalone configuration-path handling in assemble,
+      associate, and select while preserving backward compatibility
+      with direct config-file calls.
+    - Fixed top-level --comment handling so assemble comments no longer
+      interfere with chained module parsing.
+    - Improved transfer/augment robustness, including cleanup on failed
+      PDL transfers, missing-email attachments, output-directory handling,
+      packet JSON ingestion, and .saved state management.
+    - Fixed multi-component product handling in raster, contour,
+      mapping, uncertainty-map, and plot regression modules to prevent
+      wrong-component selection, filename overwrites, and contents.xml
+      collisions.
+    - Standardized preferred-component selection, using
+      GREATER_OF_TWO_HORIZONTAL when available with deterministic
+      fallback behavior.
+    - Improved mapping/KML/shape reliability without changing
+      established rendering behavior; fixed component-order dependence,
+      container cleanup, map target metadata, and several small resource
+      leaks.
+    - Preserved and integrated recent antimeridian, ocean-contour,
+      hillshade, Natural Earth, and mapping performance improvements.
+    - Improved DYFI robustness: optional retrieval failures are now
+      nonfatal throughout, GeoJSON station names are parsed correctly,
+      and dateline-crossing cells get correct center longitudes.
+    - Vectorized DYFI uncertainty lookup with identical threshold behavior
+      and a large performance improvement.
+    - Hardened history/version handling, backup parsing, empty-history
+      behavior, and HDF container cleanup.
+    - Left the refactored model orchestration layer intentionally thin;
+      no additional changes were needed after review.
+- Refactor b20
+    - Updated finite-rupture prediction code to request uncertainty
+      components from the actual filtered GMPE or IPE being evaluated,
+      rather than from cached capabilities of the unfiltered default
+      model.
+    - Removed obsolete cached uncertainty fields from PredictorSet
+      (gmpe_total_sd_only, gmpe_stddev_types, ipe_total_sd_only, and
+      ipe_stddev_types) and centralized predictor capability resolution.
+    - Changed internal tau/phi fallback logic to depend on the uncertainty
+      components actually returned by a prediction; ShakeMap synthesizes
+      a tau/phi split only when the prediction supplies total sigma alone.
+    - Added provenance to station predictions so ShakeMap can distinguish
+      predictor-supplied tau/phi from internally synthesized values.
+    - Corrected stationlist.json uncertainty output to report the
+      predictor’s total sigma directly rather than reconstructing sigma
+      from tau and phi.
+    - Changed stationlist.json to report tau and phi only when they were
+      supplied by the unconditioned prediction; synthetic internal tau/phi
+      values are now serialized as null.
+    - Added regression coverage for mixed per-IMT uncertainty capabilities,
+      filtered GMPEs, VirtualIPE-derived predictions, total-only predictions,
+      and station JSON uncertainty provenance.
+- Restored dashed ground-motion contours over water while preserving solid
+  contours over land. Replaced the previous double-draw/z-order approach
+  with direct land/water segmentation using the pre-rendered basemap
+  raster, reducing contour rendering overhead.
+- Refactor b19 - Masking and Bias
+    - Fixed custom-mask handling for point output so configured raster
+      masks are ignored rather than attempting to apply an empty grid
+      mask to point arrays.
+    - Range-limited Engler event-state conditioning.
+    - Corrected nominal-bias semantics.
+    - BiasResult/ConditioningSystem ownership cleanup.
+    - Removal of bias_max_dsigma.
+    - Corrected MMI station uncertainty selection so MMI uses the IPE
+      total-standard-deviation mode rather than the GMPE mode.
+    - Optimized land masking by clipping selected Natural Earth 10 m
+      geometries to raster pixel-edge bounds before rasterization, with
+      bit-for-bit identical masks and substantially improved performance.
+    - Fixed stale rock/soil station depth parameters by constructing
+      fresh site contexts for the rock and soil Vs30 cases.
+    - Removed the dead ModelResults.conditioning field.
+    - Made info_builder copy prediction metadata rather than mutate the
+      supplied metadata object.
+    - Cleaned up remaining directivity issues, including MMI handling and
+      stale/error-prone code.
+    - Fixed standalone model logging to use model.log rather than the
+      obsolete shape.log.
+    - Added masking-equivalence regression coverage to protect the
+      land-mask optimization.
+- Refactor b18 — Prediction/state cleanup and optimization
+    - Kept point-output arrays 1-D end-to-end and reused existing station
+      distance contexts.
+    - Reduced PredictionRuntime to rupture/directivity state and moved
+      prediction metadata construction to a pure function.
+    - Added FFSimmer/VirtualIPE source-GMPE reuse for point-rupture station
+      predictions, removing duplicate VirtualIPE GMPE evaluations.
+    - Removed hidden land-mask state and made custom masking non-mutating.
+    - Removed dead persisted fields from rupture, geometry, attenuation, and 
+      predictor products.
+    - Tested finite-rupture VirtualIPE reuse, found the speed gain too small
+      to justify the added complexity, and reverted it.
+    - Reduced representative ci3144585 model time further to about 36.5 s,
+      roughly 29% faster than b16 overall.
+- Refactor b17 — Station modeling refactor
+    - Replaced dataframe-centric station state with explicit station
+      observation, geometry, prediction, QC, bias, and group products.
+    - Removed StationDataFrame-style temporal prediction/conversion
+      bookkeeping from the modeling path.
+    - Made MMI↔PGM station conversions explicit and preserved native versus
+      derived observation provenance.
+    - Eliminated a redundant station prediction pass, cutting station FFSimmer
+      work roughly in half.
+    - Reduced representative ci3144585 model time from about 51.1 s to 38.7 s.
+- Refactor b10--b16 — Modeling architecture refactor
+    - Removed the giant ModelingContext and replaced it with explicit modeling
+      products and data flow.
+    - Broke the model workflow into focused setup, prediction, conditioning,
+      bias, MVN, and output stages.
+    - Established the architecture that b17/b18 subsequently simplified rather
+      than replacing.
+- Reconciled assemble's -p option with model.conf's 
+  interp->prediction_location->file option.
+- Broke out get_imt_type_indices function from compute_bias; added imt_type_ix 
+  array to each IMT in output HDF file.
+
 ## 1.2.2 / 2026-08-23
 - Update esi-shakelib requirement to v1.2.5.
 - Add 0.6 sec SA to default imt_list in model.conf.
 - Mapping and Uncertainty Map Improvements
-    o Hillshade and terrain rendering
+    - Hillshade and terrain rendering
         - Replaced Matplotlib LightSource.blend_hsv() with a custom RGB hillshade
           blend to avoid the dark/teal color shifts seen in some maps.
         - Retained a single directional light source for consistent terrain relief.
@@ -20,7 +141,7 @@
         - The final balance produces darker, better-defined relief, avoids the
           garish mint-green appearance in many coastal/lowland areas, and restores
           richer terrain colors.
-    o Vector-layer performance
+    - Vector-layer performance
         - Added Fiona/GDAL bbox-filtered reads for map vector layers so geometries
           outside the map region are not unnecessarily materialized.
         - Applied regional filtering to configured vector layers and Natural Earth
@@ -30,7 +151,7 @@
         - Mapping also applies regional filtering to roads and faults.
         - These changes substantially reduced Natural Earth/vector I/O and Cartopy
           geometry-projection costs.
-    o Antimeridian handling
+    - Antimeridian handling
         - Added robust detection for maps crossing 180 degrees in all supported
           longitude representations:
             * wrapped bounds, such as 179 to -179
@@ -44,7 +165,7 @@
           full-rendering path.
         - The same antimeridian behavior is used by both mapping and
           uncertaintymaps.
-    o Shared rendering and caching
+    - Shared rendering and caching
         - Reuse a single Mercator projection for all maps generated for an event.
         - Render static basemap layers once per event and reuse the resulting raster
           for all IMT maps.
@@ -54,7 +175,7 @@
           filtered/positioned city list.
         - Compute projected topography and hillshade intensity once per event and
           reuse them across regular maps.
-    o Matplotlib rendering optimizations
+    - Matplotlib rendering optimizations
         - Removed redundant fig.canvas.draw() calls when precomputed city-collision
           results are available.
         - Preserved the canvas draw fallback for callers that still require local
@@ -66,7 +187,7 @@
           before reuse.
         - Mapping and uncertainty maps can independently add small amounts of extra
           right/top padding by adjusting the cached bounding box coordinates.
-    o Uncertainty map improvements
+    - Uncertainty map improvements
         - Applied the same regional vector-reading and antimeridian optimizations as
           the regular mapping path.
         - Reused the shared projection, basemap raster, and precomputed city
@@ -80,7 +201,7 @@
         - In the profiled test case, uncertaintymaps runtime dropped from about
           68.6 seconds to about 46.4 seconds after the canvas-draw and tight-bbox
           optimizations.
-    o Performance summary
+    - Performance summary
         - The ordinary mapping path improved from roughly 56.7 seconds to roughly
           31 seconds before the later rendering optimizations, with a further modest
           improvement from eliminating redundant canvas draws and bbox calculations.
@@ -89,7 +210,7 @@
           original vector-rendering behavior.
         - Uncertainty-map rendering improved by roughly 22 seconds, or about 32
           percent, in the profiled case after eliminating duplicate rendering work.
-    o Testing and robustness
+    - Testing and robustness
         - Added regression coverage for ordinary, wrapped-antimeridian, and
           unwrapped-antimeridian clip-box construction.
         - Added tests confirming that ordinary maps use bbox-filtered reads while
@@ -99,236 +220,236 @@
     - Revised makecsv -g to provide a dummy network code of "--" for stations that
         don't have a network code.
     - Hardened station-value handling in mapmaker.py
-        o Updated map station plotting to handle Python None in addition to legacy "null"
+        - Updated map station plotting to handle Python None in addition to legacy "null"
           values.
-        o Safely handles missing/invalid observed intensity, converted instrumental intensity,
+        - Safely handles missing/invalid observed intensity, converted instrumental intensity,
           and IMT amplitudes.
-        o Missing instrumental amplitudes are skipped rather than interpreted as zero.
-        o Fixed the Tohoku mapping failure caused by float(None).
+        - Missing instrumental amplitudes are skipped rather than interpreted as zero.
+        - Fixed the Tohoku mapping failure caused by float(None).
     - Resolved high-latitude/polar mapping failures
-        o Tracked several Cartopy failures for the Franz Josef Land event through projection
+        - Tracked several Cartopy failures for the Franz Josef Land event through projection
           construction, extent setting, and raster rendering.
-        o Confirmed that using event latitude bounds as
+        - Confirmed that using event latitude bounds as
           Mercator(min_latitude=..., max_latitude=...) was incorrect.
-        o Rejected Azimuthal Equidistant and Plate Carrée alternatives after testing exposed
+        - Rejected Azimuthal Equidistant and Plate Carrée alternatives after testing exposed
           curved raster-domain artifacts and severe map distortion.
-        o Kept the existing Mercator architecture but expanded its fixed projection validity
+        - Kept the existing Mercator architecture but expanded its fixed projection validity
           range to approximately ±89.5°.
-        o Centralized Mercator construction so mapping.py, mapmaker.py, and MercatorMap all use
+        - Centralized Mercator construction so mapping.py, mapmaker.py, and MercatorMap all use
           the same projection.
-        o Restored correct map geometry, aspect, raster alignment, contours, graticules, and
+        - Restored correct map geometry, aspect, raster alignment, contours, graticules, and
           epicenter placement for the far-north event.
-        o Identified the corresponding uncertaintymaps projection update
-        o Confirmed that uncertaintymaps.py independently constructs the old default Mercator
+        - Identified the corresponding uncertaintymaps projection update
+        - Confirmed that uncertaintymaps.py independently constructs the old default Mercator
           projection.
-        o Determined that it should use the same centralized get_map_projection() helper as
+        - Determined that it should use the same centralized get_map_projection() helper as
           mapping.py.
     - Hardened station-value handling in kml.py
-        o Fixed KML generation failure when station amplitudes contain Python None.
-        o Added a shared finite-number conversion helper accepting numeric values while rejecting
+        - Fixed KML generation failure when station amplitudes contain Python None.
+        - Added a shared finite-number conversion helper accepting numeric values while rejecting
           None, "null", malformed values, NaN, and Inf.
-        o Updated get_imt_text() to ignore missing amplitudes and preserve the existing
+        - Updated get_imt_text() to ignore missing amplitudes and preserve the existing
           "nan <units>" fallback.
-        o Updated get_intensity() to safely fall back through station intensity, MMI-channel
+        - Updated get_intensity() to safely fall back through station intensity, MMI-channel
           intensity, and finally zero.
-        o Added regression tests for both None and legacy "null" values.
-        o Verified the previously failing usp000gvtu KML event now completes successfully.
+        - Added regression tests for both None and legacy "null" values.
+        - Verified the previously failing usp000gvtu KML event now completes successfully.
     - Correctness pass over model modules:
-        o Fixed station-trimming longitude padding and improved dateline/polar handling.
-        o Replaced global NumPy error-state changes with local np.errstate() contexts.
-        o Fixed mechanism/STREC precedence and kept rupture/origin context consistent.
-        o Made input/output container cleanup exception-safe.
-        o Tightened handling of optional StationList/STREC container data so unrelated AttributeErrors
+        - Fixed station-trimming longitude padding and improved dateline/polar handling.
+        - Replaced global NumPy error-state changes with local np.errstate() contexts.
+        - Fixed mechanism/STREC precedence and kept rupture/origin context consistent.
+        - Made input/output container cleanup exception-safe.
+        - Tightened handling of optional StationList/STREC container data so unrelated AttributeErrors
           are not hidden.
-        o Made finite-rupture context shape restoration exception-safe.
-        o Removed a duplicate sites_obj_out declaration from ModelingContext.
-        o Hardened stddev handling so returned arrays are associated with the exact requested stddev
+        - Made finite-rupture context shape restoration exception-safe.
+        - Removed a duplicate sites_obj_out declaration from ModelingContext.
+        - Hardened stddev handling so returned arrays are associated with the exact requested stddev
           types, with canonical ShakeMap ordering preserved.
-        o Added regression tests for nonstandard stddev ordering and validation.
+        - Added regression tests for nonstandard stddev ordering and validation.
     - Streamline build-test-deploy pipeline in .gitlab-ci.yml.
     - Fix mapping bug reported by Anthony C. of GA.
     - Cleanup/correctness pass on scalebar.py.
     - Cleanup of codes in shakemap_modules/base:
-        o Cleaned and simplified CLI argument handling
-            o Restored optional logdir and transfer outdir behavior to match the documented CLI semantics.
-            o Simplified path validation and error handling with pathlib and parser.error().
-            o Limited returned custom-argument dictionaries to only caller-defined options.
-        o Improved CoreModule/contents handling
-            o Fixed cross-platform contents.xml download paths to always use portable forward slashes.
-            o Hardened XML parsing for empty captions and unexpected elements.
-            o Simplified filesystem and XML writing logic.
-        o Refined transfer base infrastructure
-            o Modernized path and UTC handling.
-            o Simplified backup discovery/versioning and product-property updates.
-            o Preserved existing transfer-blocking, backup, review-status, and argument-parsing behavior.
-        o General maintenance
-            o Removed redundant code and older path-handling patterns.
+        - Cleaned and simplified CLI argument handling
+            - Restored optional logdir and transfer outdir behavior to match the documented CLI semantics.
+            - Simplified path validation and error handling with pathlib and parser.error().
+            - Limited returned custom-argument dictionaries to only caller-defined options.
+        - Improved CoreModule/contents handling
+            - Fixed cross-platform contents.xml download paths to always use portable forward slashes.
+            - Hardened XML parsing for empty captions and unexpected elements.
+            - Simplified filesystem and XML writing logic.
+        - Refined transfer base infrastructure
+            - Modernized path and UTC handling.
+            - Simplified backup discovery/versioning and product-property updates.
+            - Preserved existing transfer-blocking, backup, review-status, and argument-parsing behavior.
+        - General maintenance
+            - Removed redundant code and older path-handling patterns.
     - Refactor of files in shakemap_modules/utils:
-        o Improved configuration handling and validation
-            o Cleaned up config-path handling, error messages, logging, and configspec usage.
-            o Preserved existing validator semantics while modernizing implementation details.
-            o Added clearer failure behavior for invalid or incomplete configuration states.
-        o Hardened model and GMPE selection utilities
-            o Added protection against invalid/nonfinite tectonic-region probabilities and STREC
+        - Improved configuration handling and validation
+            - Cleaned up config-path handling, error messages, logging, and configspec usage.
+            - Preserved existing validator semantics while modernizing implementation details.
+            - Added clearer failure behavior for invalid or incomplete configuration states.
+        - Hardened model and GMPE selection utilities
+            - Added protection against invalid/nonfinite tectonic-region probabilities and STREC
               distances.
-            o Clarified slab-model handling and subduction GMPE selection.
-            o Preserved existing probability/taper formulas while tightening validation around
+            - Clarified slab-model handling and subduction GMPE selection.
+            - Preserved existing probability/taper formulas while tightening validation around
               unsupported configurations.
-        o Improved station/data conversion utilities
-            o Made station/point conversion more deterministic and robust to malformed IDs, missing IMTs,
+        - Improved station/data conversion utilities
+            - Made station/point conversion more deterministic and robust to malformed IDs, missing IMTs,
               and incomplete data.
-            o Improved temporary-file cleanup, SQLite handling, and JSON/XML serialization.
-            o Standardized UTF-8 handling and safer ASCII normalization for problematic Unicode station names.
-        o Strengthened amplitude database handling
-            o Fixed duplicate-station association so each candidate uses its own travel-time estimate.
-            o Corrected channel matching to distinguish identical channel codes with different location codes.
-            o Parameterized SQL, improved transaction/resource handling, and corrected timezone conversion
+            - Improved temporary-file cleanup, SQLite handling, and JSON/XML serialization.
+            - Standardized UTF-8 handling and safer ASCII normalization for problematic Unicode station names.
+        - Strengthened amplitude database handling
+            - Fixed duplicate-station association so each candidate uses its own travel-time estimate.
+            - Corrected channel matching to distinguish identical channel codes with different location codes.
+            - Parameterized SQL, improved transaction/resource handling, and corrected timezone conversion
               behavior.
-        o Improved ShakeMap cloning support
-            o Fixed generated model.conf details, including outlier settings and independent X/Y grid spacing.
-            o Added HTTP timeouts/status checking and reliable temporary-file cleanup.
-            o Improved station filtering and database cleanup while retaining historical ShakeMap migration
+        - Improved ShakeMap cloning support
+            - Fixed generated model.conf details, including outlier settings and independent X/Y grid spacing.
+            - Added HTTP timeouts/status checking and reliable temporary-file cleanup.
+            - Improved station filtering and database cleanup while retaining historical ShakeMap migration
               behavior.
-        o Improved logging, layers, amplification, and dataframe helpers
-            o Fixed logging handler configuration and level handling.
-            o Corrected polygon edge-distance handling for closing segments.
-            o Hardened generic amplification grid access and metadata handling.
-            o Simplified dataframe/XML conversion and removed unused intermediate database work.
-        o Removed obsolete and dead code
-            o Removed unused imports, empty methods, redundant configuration plumbing, legacy helper code, and
+        - Improved logging, layers, amplification, and dataframe helpers
+            - Fixed logging handler configuration and level handling.
+            - Corrected polygon edge-distance handling for closing segments.
+            - Hardened generic amplification grid access and metadata handling.
+            - Simplified dataframe/XML conversion and removed unused intermediate database work.
+        - Removed obsolete and dead code
+            - Removed unused imports, empty methods, redundant configuration plumbing, legacy helper code, and
               decorative/commented-out blocks.
-            o Replaced unnecessary regexes and older Python idioms with simpler modern equivalents.
-        o General maintenance
-            o Standardized lazy logging, context-managed resources, explicit UTF-8 handling, modern path handling,
+            - Replaced unnecessary regexes and older Python idioms with simpler modern equivalents.
+        - General maintenance
+            - Standardized lazy logging, context-managed resources, explicit UTF-8 handling, modern path handling,
               deterministic ordering, and clearer exceptions.
     - Coremods: assemble augment coverage dyfi gridxml info makecsv raster rupture save shape
       sleep sm_select stations transfer_email transfer_pdl transfer_scp:
-        o Resource safety: Reworked many modules so HDF containers, ZIP files, SQLite connections,
+        - Resource safety: Reworked many modules so HDF containers, ZIP files, SQLite connections,
           and ordinary files are reliably closed with try/finally or context managers, including
           failure paths.
-        o Safer failure behavior: Delayed destructive cleanup of old products/backups until new
+        - Safer failure behavior: Delayed destructive cleanup of old products/backups until new
           processing had succeeded, removed inappropriate ignore_errors=True, and added clearer
           standalone argument/directory validation.
-        o Dependency and target correctness: Fixed stale or incorrect dependency declarations, added
+        - Dependency and target correctness: Fixed stale or incorrect dependency declarations, added
           missing optional inputs and generated targets, and improved handling of dynamically named products.
-        o Multi-component correctness: Removed assumptions that only the first IMC/component mattered,
+        - Multi-component correctness: Removed assumptions that only the first IMC/component mattered,
           preserved legacy filenames for GREATER_OF_TWO_HORIZONTAL, and added component-specific
           naming/contents where needed.
-        o Metadata correctness: Stopped relying on incidental “last IMT processed” state, used each
+        - Metadata correctness: Stopped relying on incidental “last IMT processed” state, used each
           layer’s own metadata where appropriate, and added checks for consistent grid geometry.
-        o Data integrity fixes: Corrected several subtle problems such as reversed full-grid latitude
+        - Data integrity fixes: Corrected several subtle problems such as reversed full-grid latitude
           ordering, cross-component dataframe column leakage, incorrect DYFI standard-deviation behavior,
           and per-destination transfer state leaking between destinations.
-        o Configuration robustness: Standardized str(get_configspec(...)) usage for ConfigObj, validated
+        - Configuration robustness: Standardized str(get_configspec(...)) usage for ConfigObj, validated
           merged configurations rather than only globals, and made standalone configuration handling more
           explicit.
-        o Transfer-system cleanup: Fixed NO_TRANSFER control flow, made Email honor mail_once, corrected
+        - Transfer-system cleanup: Fixed NO_TRANSFER control flow, made Email honor mail_once, corrected
           per-PDL-destination system_id handling, avoided mutating shared ConfigObj sections, and made
           temporary transfer directories exception-safe.
-        o Deterministic behavior: Sorted glob results where file ordering could affect which input was
+        - Deterministic behavior: Sorted glob results where file ordering could affect which input was
           selected, and hardened backup-version parsing.
-        o Logging cleanup: Converted many messages to lazy logger formatting, fixed copied/wrong logfile names
+        - Logging cleanup: Converted many messages to lazy logger formatting, fixed copied/wrong logfile names
           and messages, replaced deprecated logging calls, and standardized use of module loggers.
-        o Modern Python cleanup: Replaced old-style super(...), removed dead imports/constants/scaffolding,
+        - Modern Python cleanup: Replaced old-style super(...), removed dead imports/constants/scaffolding,
           simplified redundant subclasses and branches, used timezone-aware UTC timestamps, and switched file
           I/O to explicit UTF-8.
-        o API consistency: Fixed mismatches between constructors, execute() methods, and standalone main()
+        - API consistency: Fixed mismatches between constructors, execute() methods, and standalone main()
           calls, and initialized module state consistently.
-        o Code quality without gratuitous behavioral change: Removed a lot of dead or misleading code, kept
+        - Code quality without gratuitous behavioral change: Removed a lot of dead or misleading code, kept
           lines within 79 columns, and generally preserved scientific/product behavior unless there was a
           clear correctness bug or confirmed intended change.
-        o Overall, the common theme was making the coremods more predictable, failure-safe, deterministic, and
+        - Overall, the common theme was making the coremods more predictable, failure-safe, deterministic, and
           internally consistent while also deleting quite a bit of historical cruft.
     - assemble.py:
-        o expanded dependencies to include the event-local inputs that actually affect assemble
-        o fixed backup selection to sort backupN directories numerically instead of lexicographically
-        o corrected simulation.conf validation errors to report against sim_config
-        o ensured all ShakeMapInputContainer instances are closed with try/finally
-        o changed history.json reading to a context manager
-        o added explicit validation for standalone config, indir, and outdir
-        o removed ignore_errors=True from destructive directory cleanup
-        o fixed the stale .transferred comment to .saved
-        o normalized get_configspec("simulation") with str(...)
-        o made X/Y simulation reshaping honor row/column ordering
-        o strengthened LAT/LON and X/Y regular-grid validation
-        o added explicit errors for missing coordinate columns and incorrect simulation row counts
-        o convert nonfinite/nonpositive simulated amplitudes to NaN before taking logs
-        o hardened _trim_grid() against all-invalid and untrimmable grids
-        o changed the standalone logfile from the apparent copy/paste shape.log to assemble.log
-        o modernized super(), set literals, truthiness checks, and some logging/style details
-        o removed the leftover # ADD THIS comment
+        - expanded dependencies to include the event-local inputs that actually affect assemble
+        - fixed backup selection to sort backupN directories numerically instead of lexicographically
+        - corrected simulation.conf validation errors to report against sim_config
+        - ensured all ShakeMapInputContainer instances are closed with try/finally
+        - changed history.json reading to a context manager
+        - added explicit validation for standalone config, indir, and outdir
+        - removed ignore_errors=True from destructive directory cleanup
+        - fixed the stale .transferred comment to .saved
+        - normalized get_configspec("simulation") with str(...)
+        - made X/Y simulation reshaping honor row/column ordering
+        - strengthened LAT/LON and X/Y regular-grid validation
+        - added explicit errors for missing coordinate columns and incorrect simulation row counts
+        - convert nonfinite/nonpositive simulated amplitudes to NaN before taking logs
+        - hardened _trim_grid() against all-invalid and untrimmable grids
+        - changed the standalone logfile from the apparent copy/paste shape.log to assemble.log
+        - modernized super(), set literals, truthiness checks, and some logging/style details
+        - removed the leftover # ADD THIS comment
     - shape.py:
-        o Fixed AREA/PERIMETER warnings.
-        o General clean up of module code.
+        - Fixed AREA/PERIMETER warnings.
+        - General clean up of module code.
     - Suppressed a GDAL warning emitted by rasterio.features.shapes() in mapping.py.
     - Changed the name of logging.py to logging_utils.py. Updated all affected modules.
     - Modified config.py:
-        o Improved configuration logging by replacing deprecated logging.warn() calls with
+        - Improved configuration logging by replacing deprecated logging.warn() calls with
           logging.warning() and converting logging messages to lazy formatting.
-        o Added explicit exception chaining when validators convert low-level TypeError/ValueError
+        - Added explicit exception chaining when validators convert low-level TypeError/ValueError
           exceptions into ValidateError, preserving the original cause for debugging.
-        o Simplified and hardened annotatedfloat_type() by removing broad exception handling
+        - Simplified and hardened annotatedfloat_type() by removing broad exception handling
           and validating supported unit suffixes explicitly.
-        o Fixed cfg_bool() so native boolean values are handled correctly and invalid boolean
+        - Fixed cfg_bool() so native boolean values are handled correctly and invalid boolean
           strings raise ValidateError instead of being silently treated as False.
-        o Fixed check_all_configs() so a validation exception cannot leave results undefined
+        - Fixed check_all_configs() so a validation exception cannot leave results undefined
           before later use.
-        o Fixed check_profile_config() to avoid deleting profile entries while iterating over
+        - Fixed check_profile_config() to avoid deleting profile entries while iterating over
           the live dictionary key view.
-        o Improved validator robustness for malformed or unexpected input types while preserving
+        - Improved validator robustness for malformed or unexpected input types while preserving
           the existing configuration-validation behavior where appropriate.
-        o General cleanup for pylint compliance and clearer error reporting without changing valid
+        - General cleanup for pylint compliance and clearer error reporting without changing valid
           ShakeMap configuration semantics.
     - Optimized KML generation
-        o Reduced KML runtime from roughly 5.5 s to ~3 s.
-        o Improved contour-generation efficiency while retaining the existing Natural Earth 10m ocean-mask behavior.
-        o Removed dead constants, unused legacy functions, redundant code, and stale documentation.
-        o Improved NaN handling in MMI polygon generation.
+        - Reduced KML runtime from roughly 5.5 s to ~3 s.
+        - Improved contour-generation efficiency while retaining the existing Natural Earth 10m ocean-mask behavior.
+        - Removed dead constants, unused legacy functions, redundant code, and stale documentation.
+        - Improved NaN handling in MMI polygon generation.
     - Fixed KML layer visibility and organization
-        o Corrected station, contour, and polygon visibility behavior in Google Earth by removing redundant
+        - Corrected station, contour, and polygon visibility behavior in Google Earth by removing redundant
           child-level visibility=0 settings.
-        o Moved individual IMT contour folders to the top level instead of nesting them under a common Contours folder.
-        o Verified that the missing intensity legend in Google Earth Web is due to unsupported ScreenOverlay
+        - Moved individual IMT contour folders to the top level instead of nesting them under a common Contours folder.
+        - Verified that the missing intensity legend in Google Earth Web is due to unsupported ScreenOverlay
           rendering rather than bad KML or a missing/corrupt legend image.
-        o Retained a conventional ScreenOverlay implementation for clients that support it.
-        o Updated KML tests for the revised hierarchy and improved their resource/error handling.
+        - Retained a conventional ScreenOverlay implementation for clients that support it.
+        - Updated KML tests for the revised hierarchy and improved their resource/error handling.
     - Cleaned coremods/contour.py
-        o Removed redundant directory checks and dead constructor filter state.
-        o Standardized contour filter sizing on products.conf.
-        o Ensured the ShakeMap output container is always closed.
-        o Replaced manual Fiona/JSON resource management with context managers.
-        o Removed overly broad GMICE exception handling so invalid configuration now fails explicitly.
-        o Clarified GMICE applicability logic for MMI, PGA/PGV, and SA.
-        o Consolidated duplicate schemas.
-        o Corrected empty-contour GeoJSON bbox output to use a numeric array rather than a string.
-        o Cleaned variable shadowing, comments, docstrings, and minor redundancies.
-        o Preserved required conversion of configspec paths to strings for ConfigObj.
+        - Removed redundant directory checks and dead constructor filter state.
+        - Standardized contour filter sizing on products.conf.
+        - Ensured the ShakeMap output container is always closed.
+        - Replaced manual Fiona/JSON resource management with context managers.
+        - Removed overly broad GMICE exception handling so invalid configuration now fails explicitly.
+        - Clarified GMICE applicability logic for MMI, PGA/PGV, and SA.
+        - Consolidated duplicate schemas.
+        - Corrected empty-contour GeoJSON bbox output to use a numeric array rather than a string.
+        - Cleaned variable shadowing, comments, docstrings, and minor redundancies.
+        - Preserved required conversion of configspec paths to strings for ConfigObj.
     - Major optimization of plotregr.py
-        o Reduced normal runtime from roughly 4.5 s to ~0.5 s.
-        o Reduced profiled runtime from 8.4 s to ~1.1 s.
-        o Replaced thousands of per-station Matplotlib calls with batched seismic and macroseismic station plotting.
-        o Reduced Matplotlib object creation and eliminated most of the expensive legend-placement overhead.
-        o Fixed a multi-component bug where the output container was closed inside the component loop.
-        o Reused attenuation arrays instead of rereading them for JSON output.
-        o Reduced data passed to plotting routines and cleaned file/resource handling.
-        o Removed obsolete multiprocessing support after benchmarking showed 8 workers increased runtime from ~0.5 s to ~4.5 s.
-        o Simplified plotregr to always render serially.
+        - Reduced normal runtime from roughly 4.5 s to ~0.5 s.
+        - Reduced profiled runtime from 8.4 s to ~1.1 s.
+        - Replaced thousands of per-station Matplotlib calls with batched seismic and macroseismic station plotting.
+        - Reduced Matplotlib object creation and eliminated most of the expensive legend-placement overhead.
+        - Fixed a multi-component bug where the output container was closed inside the component loop.
+        - Reused attenuation arrays instead of rereading them for JSON output.
+        - Reduced data passed to plotting routines and cleaned file/resource handling.
+        - Removed obsolete multiprocessing support after benchmarking showed 8 workers increased runtime from ~0.5 s to ~4.5 s.
+        - Simplified plotregr to always render serially.
     - Fixed bug in logging.py.
     - Tried to make multiprocessing of mapping and plotting modules a little safer.
     - Removed multiprocessing from shape.py because it made it slower.
     - Remove annoying 'Down-casting double precision floating point to single precision' warning in raster.py.
     - Removed obsolete numpy RankWarning monkeypatch.
     - mapping and uncertaintymaps optimizations
-        o Cache city-label collision detection once per event instead of recomputing it for every map.
-        o Switch station-marker plotting from per-point ax.plot() calls to batched ax.scatter() (both filled
+        - Cache city-label collision detection once per event instead of recomputing it for every map.
+        - Switch station-marker plotting from per-point ax.plot() calls to batched ax.scatter() (both filled
           and unfilled variants) — fixes a rendering hang as well as performance.
-        o Deduplicate contour-line geometry projection (was projecting the same contours twice, once each for
+        - Deduplicate contour-line geometry projection (was projecting the same contours twice, once each for
           the dashed and solid passes).
-        o Skip topography reprojection entirely in draw_uncertainty_map() (result was computed but never used);
+        - Skip topography reprojection entirely in draw_uncertainty_map() (result was computed but never used);
           cache the projected topography + hillshade once per event in draw_map() instead of recomputing per IMT.
-        o Add explicit transform=ax.transData to per-map data-image imshow calls to avoid unnecessary Cartopy CRS re-warping.
-        o Extend all of the above (basemap raster, collision caching) to uncertaintymaps.py, which hadn't had any of it.
-        o Fix xmin/xmax/ymin/ymax not being cast to float in uncertaintymaps.py, causing a TypeError in GeoDict.createDictFromBox.
+        - Add explicit transform=ax.transData to per-map data-image imshow calls to avoid unnecessary Cartopy CRS re-warping.
+        - Extend all of the above (basemap raster, collision caching) to uncertaintymaps.py, which hadn't had any of it.
+        - Fix xmin/xmax/ymin/ymax not being cast to float in uncertaintymaps.py, causing a TypeError in GeoDict.createDictFromBox.
 
 ## 1.2.1 / 2026-08-14
     - Add blas_threads and openmp_threads config parameters to model.conf.
@@ -337,15 +458,15 @@
 
 ## 1.2.0 / 2026-08-10
     - Optimize mapping performance:
-        o Render the static basemap/faults/roads layers once instead of once per IMT map
-        o Vectorize the station plotting
-        o Deduplicate the contour reprojection
-        o Clip faults/roads to the map bounds before handing them to Cartopy
+        - Render the static basemap/faults/roads layers once instead of once per IMT map
+        - Vectorize the station plotting
+        - Deduplicate the contour reprojection
+        - Clip faults/roads to the map bounds before handing them to Cartopy
     - Refactor model:
-        o Refactor model.py into a suite of smaller modules.
-        o Create unit tests for all of the new modules.
-        o Optimize performance of model.
-        o Remove max_workers config parameter; add row_block.
+        - Refactor model.py into a suite of smaller modules.
+        - Create unit tests for all of the new modules.
+        - Optimize performance of model.
+        - Remove max_workers config parameter; add row_block.
     - Make the RNG seed for FFSimmer a function of event ID instead of a constant.
     - Change esi-core>=1.2.8, gmpacket>=0.1.10, and esi-shakelib>=1.2.2 in pyproject.toml.
 

{shakemap_modules-1.2.2 → shakemap_modules-1.2.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: shakemap-modules
-Version: 1.2.2
+Version: 1.2.3
 Summary: USGS Earthquake Impact ShakeMap Modules
 Author-email: Bruce Worden <cbworden@contractor.usgs.gov>, Eric Thompson <emthompsone@usgs.gov>, Mike Hearne <mhearne@usgs.gov>
 License-Expression: CC0-1.0
@@ -8,7 +8,7 @@ Requires-Python: <3.13,>=3.12
 Description-Content-Type: text/markdown
 License-File: LICENSE.md
 Requires-Dist: configobj>=5.0.8
-Requires-Dist: esi-shakelib>=1.2.5
+Requires-Dist: esi-shakelib>=1.2.6
 Requires-Dist: h5py>=3.8.0
 Requires-Dist: importlib-resources>=6.0.0
 Provides-Extra: assemble