BLOB will store backslash in name, but can't retrieve data

[msweier]

I can store a blob like PLOTS/HISTORICAL/LOCKDAM_10.FLOW-OUT.JSON, but I can't retrieve the data stored within it out of CDA. I suspect the / is the issue.

Expected behavior
An entry that can be viewed here:
curl -X 'GET' \ 'https://cwms-data.usace.army.mil/cwms-data/blobs?office=MVP&like=%2FPLOTS%2FHISTORICAL%2FLOCKDAM_10.FLOW-OUT.JSON' \ -H 'accept: application/json'
which returns
{ "blobs": [ { "office-id": "MVP", "id": "/PLOTS/HISTORICAL/LOCKDAM_10.FLOW-OUT.JSON", "description": "Historical Flow-Out Plot: Daily Values Compared to Historical Percentiles. Generated: 2026-01-22 09:20:14+00:00 UTC", "media-type-id": "application/json" } ], "page": "fHwwfHwyMA==", "page-size": 20, "total": 0 }

but when I try and retrieve the data stored under that entry with
curl -X 'GET' \ 'https://cwms-data.usace.army.mil/cwms-data/blobs/%2FPLOTS%2FHISTORICAL%2FLOCKDAM_10.FLOW-OUT.JSON?office=MVP' \ -H 'accept: */*'

a 404 Not Found message is returned.

I can retrieve the data if I store it asplots-historical-LockDam_10.Flow-Out.json.

CDA Version (please complete the following information):
CDA version 2025.09.02

[krowvin]

Possibly resolved with #1483 ?

Should be able to test in Dev

The text was updated for blob-id to say

If this query parameter is provided the id path parameter is ignored and the value of the query parameter is used. Note: this query parameter is necessary for id's that contain '/' or other special characters. This is due to limitations in path pattern matching. We will likely add support for encoding the ID in the path in the future. For now use the id field for those IDs. Client libraries should detect slashes and choose the appropriate field. "ignored" is suggested for the path endpoint.

Mitch you might see if you are able to POST/GET on dev with your blob path? Extra testing before that goes prod would be great!

I also mentioned to @MikeNeilson that i'm not sure why we have to use /blobs/ignored?blob-id instead of just /blobs?blob-id and catch when blob-id is provided to the getAll method? Might be a limitation of Javalin?

[MikeNeilson]

Might be a limitation of Javalin?

I think @DanielTOsborne just didn't want to tweak the implementation within the getAll. Can't say I blame him (and I certainly didn't even noticed that when I reviewed the PR) though we do it in several other handlers.

Considering the importance of the Blob endpoint I think this is a prime candidate for "let's make a v2" that had far better agreed upon semantics.

[rma-rripken]

The reason for the "ignored" in /blobs/ignored?blob-id is so that the request gets routed by javalin to the right controller end-point. The controller methods are registered linking a url template to a method. You want your request to hit the getOne method and not the getAll method. There is another method in the BlobController that can return a list of blobs, that method doesn't take an id in the path and would be where your request should be routed if there isn't something that looks like an id in the path. Making the return type of the getAll method conditional on whether or not a blob-id parameter was provided seemed gross and difficult to do in OpenAPI. If we have the blob-id query parameter override the id path parameter then everything works as long as the path parameter doesn't contain chars that tomcat doesn't like. A while back I suggested "ignored" was used but it isn't required - it works as long as something valid is there. I'm not saying all of this is ideal - just trying to explain the backstory.

[msweier]

Mitch you might see if you are able to POST/GET on dev with your blob path? Extra testing before that goes prod would be great!

I am able to post a very small test blob, but I'm not able to GET. I'm not sure I'm requesting this correctly though. There's two blob-id arguments.

curl -X 'GET' \
  'https://DEV/cwms-data/blobs/%2FPLOTS%2FHISTORICAL%2FLOCKDAM_10.FLOW?office=MVP&blob-id=%2FPLOTS%2FHISTORICAL%2FLOCKDAM_10.FLOW&_cb=1769113937800' \
  -H 'accept: application/octet-stream' \
  -H 'Cache-Control: no-cache, no-store, max-age=0' \
  -H 'Pragma: no-cache'

[MikeNeilson]

A while back I suggested "ignored" was used but it isn't required - it works as long as something valid is there. I'm not saying all of this is ideal - just trying to explain the backstory.

Ah, that explains it. In hindsight, perhaps we should've made a more explicit end point like /blob/by-query-param or something.

[MikeNeilson]

Mitch, try putting the ignored in the path portion (before the ?) and leave the other id.

[msweier]

Still getting a 404 for these too...not sure the magic combo:

/cwms-data/blobs/%2FPLOTS%2FHISTORICAL%2FLOCKDAM_10.FLOW?office=MVP&blob-id=%2FPLOTS%2FHISTORICAL%2F&_cb=1769117533061'
 

/cwms-data/blobs/%2FPLOTS%2FHISTORICAL%2F?office=MVP&blob-id=LOCKDAM_10.FLOW&_cb=1769117710207' 

[msweier]

Well an unforeseen side effect is you can't delete ids with a "/" in them either. Makes sense since you can't read them.