re-initialize TODOs with new action

.deployment/tofu/modules/titiler/main.tf

@@ -98,7 +98,7 @@ resource "google_artifact_registry_repository" "burn-backend" {
   location      = "us-central1"
 }
 
-## TODO: Both this and the burn-backend service should either be require authenticated
+## TODO(!feat): Both this and the burn-backend service should either be require authenticated
 ## invocations or at least have a rate limit
 
 # Allow unauthenticated invocations

src/burn_backend/lib/derive_boundary.py

@@ -269,7 +269,7 @@ def derive_boundary(
     pipeline,
 ):
 
-    ## TODO: Some part of the spectral index process is creating a buffer of NaN
+    ## TODO(!smelly): Some part of the spectral index process is creating a buffer of NaN
     ## at the outside edge of the metric layer - not an issue to replace with 0 in this case
     ## but zeros inside the image will be erroneously identified as unburned islands which is
     ## a big problem.
@@ -279,11 +279,10 @@ def derive_boundary(
         interior_nan_filled, metric_values_exist_binary
     )
 
-    # TODO: This is a new issue that should be picked up by the github action todo-to-issue
-    # Test test, hello world
-
-    # TODO(!urgent): This is a new issue that should be picked up by the github action todo-to-issue and its real serious!
-    # Issue URL: https://github.com/SchmidtDSE/burn-severity-mapping-poc/issues/48
+    # TODO (!feat): Improve (and investigate) handling of internal NaNs within derived boundary
+    # Internal NaNs are a problem because they may be interpreted as unburned islands, unless we handle
+    # them directly. So far so good, it appears we only get these at the edges of the boundary where we
+    # may have interpolation issues, so this is very conservative, but a little hacky.
 
     if no_interior_nan_detected:
         # In this case, we aren't missing interior unburned islands, but we still want the original

src/burn_backend/lib/query_sentinel.py

@@ -73,7 +73,7 @@ def __init__(
         self.crs = crs
         self.buffer = buffer
 
-        # TODO [#17]: Settle on standards for storing polygons
+        # TODO(!feat): Settle on standards for storing polygons
         # Oscillating between geojsons and geopandas dataframes, which is a bit messy. Should pick one and stick with it.
         self.geojson_boundary = None
         self.bbox = None
@@ -99,7 +99,7 @@ def set_boundary(self, geojson_boundary):
             None
         """
         boundary_gpd = gpd.GeoDataFrame.from_features(geojson_boundary)
-        # TODO [#7]: Generalize Sentinel2Client to accept any CRS
+        # TODO(!feat): Generalize Sentinel2Client to accept any CRS
         # This is hard-coded to assume 4326 - when we draw an AOI, we will change this logic depending on what makes frontend sense
         if not boundary_gpd.crs:
             geojson_boundary = boundary_gpd.set_crs("EPSG:4326")
@@ -147,7 +147,7 @@ def get_items(self, date_range, from_bbox=True, max_items=None):
         """
         date_range_fmt = "{}/{}".format(date_range[0], date_range[1])
 
-        # TODO [#14]: Cloud cover response to smoke
+        # TODO(!feat): Cloud cover response to smoke
         # Right now we don't give any mind to smoke occlusion, but we should considering we will have bias if smoke occludes our imagery
 
         query = {
@@ -266,7 +266,7 @@ def reduce_time_range(self, range_stack):
             xarray.DataArray: The reduced range stack.
         """
 
-        # TODO [#30]: Think about best practice for reducing time dimension pre/post fire
+        # TODO(!feat): Think about best practice for reducing time dimension pre/post fire
         # This will probably get a bit more sophisticated, but for now, just take the median
         # We will probably run into issues of cloud occlusion, and for really long fire events,
         # we might want to look into time-series effects of greenup, drying, etc, in the adjacent
@@ -426,7 +426,7 @@ def derive_boundary_flood_fill(
 
             seed_indices = list(zip(*np.where(metric_layer["seed"].values[0, :, :])))
 
-        ## TODO: Seed indices are essentially required at the moment, but this is an artifact
+        ## TODO(!smelly): Seed indices are essentially required at the moment, but this is an artifact
         ## of flood fill segmentation, so this Pipeline should be more flexible in the future.
         pipeline = Pipeline(
             thresholding_strategy=OtsuThreshold(),

src/burn_backend/routers/analyze/spectral_burn_metrics.py

@@ -44,7 +44,7 @@ class AnaylzeBurnPOSTBody(BaseModel):
     final: bool = True
 
 
-# TODO [#5]: Decide on / implement cloud tasks or other async batch
+# TODO(!feat): Decide on / implement cloud tasks or other async batch
 # This is a long running process, and users probably don't mind getting an email notification
 # or something similar when the process is complete. Esp if the frontend remanins static.
 @router.post(

src/burn_backend/routers/batch/batch_analyze_and_fetch.py

@@ -140,7 +140,6 @@ def main(
     }
     geojson_name = "drawn_aoi_boundary" if derive_boundary else "boundary"
 
-    ## TODO [#34]: Should probably define a class for batch analysis and fetch
     job_status = {
         "submitted": str(submission_time),
         "fire_event_name": fire_event_name,

src/burn_backend/routers/refine/flood_fill_segmentation.py

@@ -38,7 +38,7 @@ class FloodFillSegmentationPOSTBody(BaseModel):
     affiliation: str
 
 
-# TODO [#5]: Decide on / implement cloud tasks or other async batch
+# TODO(!feat): Decide on / implement cloud tasks or other async batch
 # This is a long running process, and users probably don't mind getting an email notification
 # or something similar when the process is complete. Esp if the frontend remanins static.
 @router.post(
@@ -66,7 +66,9 @@ async def refine_flood_fill_segmentation(
     """
     sentry_sdk.set_context("fire-event", {"request": body})
 
-    # TODO: No idea where this is getting converted to a dict...
+    # TODO(!smelly): geojson interpreted inconsistently as string or dict by FastAPI in burn-backend?
+    # Likely caused by a frontend issue with the request body - need to investigate
+
     # geojson_seed_points = json.loads(body.geojson)
     user_edits_geojson = body.geojson
 
@@ -89,21 +91,13 @@ async def main(
     logger,
     cloud_static_io_client,
 ):
-    ## NOTE: derive_boundary is accepted for now to maintain compatibility with the frontend,
-    ## but will shortly be a different endpoint
-
     logger.info(f"Received flood-fill-segmentation request for {fire_event_name}")
 
     try:
         # create a Sentinel2Client instance, without initializing, since we are
         # getting what we need from the cogs already generated
         geo_client = Sentinel2Client()
 
-        ## TODO: Since we are running serverless, and don't have a live database, we are
-        ## required to re-construct the metrics stack from the existing files, in the case
-        ## where the user has identified fire boundaries from our intermediate rbr output.
-        ## This is not ideal, but not sure there is a better solution without using something
-        ## like redis or another live cache.
         metric_layers = []
         for metric_name in ["nbr_prefire", "nbr_postfire", "dnbr", "rdnbr", "rbr"]:
             with tempfile.NamedTemporaryFile(suffix=".tif", delete=False) as tmp:
@@ -131,12 +125,10 @@ async def main(
             inplace=True,
         )
 
-        # save the derived boundary to the FTP server
+        # save the derived boundary to the S3 bucket
         with tempfile.NamedTemporaryFile(suffix=".geojson", delete=False) as tmp:
             tmp_geojson = tmp.name
 
-            ## TODO: This is a little circuitous but, was getting errors trying to
-            ## use .to_json() and saving the result
             geo_client.geojson_boundary.to_file(tmp_geojson, driver="GeoJSON")
             derived_boundary_json = json.load(open(tmp_geojson))
 

src/common/lib/backend_dependencies.py

@@ -97,7 +97,7 @@ def init_sentry(logger: Logger = Depends(get_cloud_logger)):
     """
     logger.info("Initializing Sentry client")
 
-    ## TODO [#28]: Move to sentry to environment variable if we keep sentry
+    ## TODO(!magic): Move to sentry to environment variable if we keep sentry
     sentry_sdk.init(
         dsn="https://3660129e232b3c796208a5e46945d838@o4506701219364864.ingest.sentry.io/4506701221199872",
         # Set traces_sample_rate to 1.0 to capture 100%

src/common/util/cloud_static_io.py

@@ -280,9 +280,6 @@ def upload_cogs(self, metrics_stack, fire_event_name, affiliation, final=True):
                     ds.build_overviews([2, 4, 8, 16, 32], Resampling.nearest)
                     ds.update_tags(ns="rio_overview", resampling="nearest")
 
-                ## TODO: Stupid solution to not having control over GDAL's cache
-                ## We need to be able to invalidate it after we crop, but we can't do that
-                ## and if we disable it entirely, performance is terrible.
                 if not final:
                     band_name = f"intermediate_{band_name}"
 
@@ -328,8 +325,7 @@ def upload_rap_estimates(self, rap_estimates, fire_event_name, affiliation):
         """
         with tempfile.TemporaryDirectory() as tmpdir:
             for band_name in rap_estimates.band.to_index():
-                # TODO [#23]: This is the same logic as in upload_cogs. Refactor to avoid duplication
-                # Save the band as a local COG
+                # TODO(!feat): Refactor upload_cogs and upload_rap_estimates to share code
                 local_cog_path = os.path.join(tmpdir, f"{band_name}.tif")
                 band_cog = rap_estimates.sel(band=band_name).rio
                 band_cog.to_raster(local_cog_path, driver="GTiff")

src/titiler/routers/pages/map.py

@@ -46,7 +46,6 @@ def serve_map(
 
     s3_bucket_name = os.getenv("S3_BUCKET_NAME")
 
-    ## TODO [#21]: Use Tofu Output to construct hardocded cog and geojson urls (in case we change s3 bucket name)
     cog_url = f"https://{s3_bucket_name}.s3.us-east-2.amazonaws.com/public/{affiliation}/{fire_event_name}/{burn_metric}.tif"
     burn_boundary_geojson_url = f"https://{s3_bucket_name}.s3.us-east-2.amazonaws.com/public/{affiliation}/{fire_event_name}/boundary.geojson"
     ecoclass_geojson_url = f"https://{s3_bucket_name}.s3.us-east-2.amazonaws.com/public/{affiliation}/{fire_event_name}/ecoclass_dominant_cover.geojson"

src/titiler/routers/pages/upload.py

@@ -20,7 +20,8 @@ def upload(
     endpoint_titiler: str = Depends(get_endpoint_titiler),
     endpoint_burn_backend: str = Depends(get_endpoint_burn_backend),
 ):
-    ## TODO: These thresholds should be configurable, and probably should use the same
+    ## TODO(!feat): Reuse UI elements from Map for Upload page
+    ## these thresholds should be configurable, and probably should use the same
     ## frontend elements as the threhsold sliders within the map. Going to punt on that for now,
     ## since the map needs a refactor in the vein of the upload refactor.
     cog_tileserver_url_prefix = (