write results into a database

[stevenstetzler]

The results are currently output into a directory, but they can instead be inserted into a database.

The current tables/schema is:

search:

• id: auto_increment int
• velocity_1: float
• velocity_2: float
• angle_1: float
• angle_2: float
• dx: float
• catalog: varchar (path)
• psfs: varchar (path)

results:

• id: auto_increment int
• search_id: foreign key to search
• x: int
• y: int
• direction: int
• n: int

tracklet:

• id: auto_increment int
• results_id: foreign key to results
• vra: float
• vdec: float
• ra_0: float
• dec_0: float
• ra_ref: float
• dec_ref: float
• tref: float
• tmin: float
• tmax: float
• sigma_vra: float
• sigma_vdec: float
• sigma_vravdec: float
• sigma_vdecvra: float
• sigma_t: float

points:

• id: auto_increment int
• results_id: foreign key to results
• ra: float
• dec: float
• time: float

gathered:

• id: auto_increment int
• results_id: foreign key to results
• ra: float
• dec: float
• time: float

gathered might be difficult, as currently (and intentionally), it's schema evolves with the input to carry metadata from the detection catalog through to the results.

points and gathered ideally would be pulled from an input database of detections + metadata through intermediate code.

[stevenstetzler]

Consider MLFlow instead of rolling your own solution