Skip to content

Allowing versioning for the dataset #320

@Bezbakri

Description

@Bezbakri

Currently the transient datasets, comprising file objects and database records, are stored without a version control system (VCS). When a transient is reprocessed, the original files are overwritten, erasing the original data files. The associated database records are more complex: new SED results, for example, are added in a non-destructive; however, there are plenty of examples currently where a particular transient has multiple SEDFittingResult objects (i.e. db records) with the same software version value. Moreover, the data displayed on the results page assumes there is only one object.

The simplest solution for the set of file objects is to simply append the current Blast version to the cutout and SED output file paths. For example, /data/cutout_cdn/2025abc/v1.2.3/ and /data/sed_output/2025abc/v1.2.3/.

We should implement a VCS to support reproducibility of output for a given Blast version and transient definition, as well as to provide (at least in principle if not by practical API) access to previous versions for analysis and comparison as needed.

Metadata

Metadata

Assignees

Labels

feature requestUse on issues that use the feature request template

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions