2021-04-10 Meeting notes

Date

10 Apr 2021

Participants

• Josh Chamberlain

• Eddie Brown

• Alec Akin

• Eric Turner

• Stabs

Goals

• Answer schema questions to unblock dolt implementation.

• Decide whether tableplus or DBeaver would be useful in the dolt pipeline (discussion here).

Discussion topics

Item	Notes
Datasets	UUIDs have been added in dolt may want to remove hyphens to save space (currently mysql built in) Once things are more stable it’s probably worth doing some views Eric wrote a script to keep CityProtect datasets up to date as well as commit the dataset to the dolt db Fork then merge is a necessity We could put these on a chron job or just manually one them—once for each portal type These update quarterly, so automating that is not our biggest problem. 355 agencies in cityprotect for bulk downloading, avg 10 CSV files We’ll need to get data out of datasets Clones and forks become problematic just to make simple additions
Dolt x Scale	Alternative for down the road: https://www.esri.com/en-us/arcgis/products/arcgis-open-data Risk: dolt is young and missing advanced SQL features. Dolt is not our advanced analytics tool. Dolt is not our deep storage layer. Databases can’t currently reference each other, and repos are 1:1 with databases. Data is unlimited but current technical cap is ~200gb Can Dolt be an intake tool and not grow? This may not even be our presentation layer because of the low amount of data it can store one way around this might be checking data out of one dolt repo into another, preserving paper trail. does dolt support this / is it reasonable? Otherwise priority ↑ for https://pdap.atlassian.net/browse/PDAP-80
Where are we storing our data properties?	Is breadth an issue in Dolt? Would this table be prohibitively gigantic? SQL column limit ~1000, mysql 4000. we should be good.
Hosting	If we went with mongodb directly we could bolt it into our DigitalOcean
Tableplus / DBeaver	context Tableplus free has some limitations (open tabs)
Scrapers	For now—people should write whatever they want / is most convenient for them. If we want to refactor, we can figure out what the priority is on it + do it later.
Server access	We don’t use passwords, just keys—send it to Alec if you need access.
OCR	Stabs tried ~5 different ones and none of them worked well enough. Alec’s tried tensorflow

Action items

• proof of concept for separate dolt DB (check whether DBs can easily reference one another) PDAP-144 Done
• proof of concept for global_properties table referenced by data_types views PDAP-121 - JIRA project doesn't exist or you don't have permission to view it.

• Josh Chamberlain to make request in dolthub discord for one:many repos:dbs

• Alec Akin reach out to 3 major cloud providers + mongodb for sponsorships

• Josh Chamberlain Point database volunteers in the direction of Dolt SQL viewer integrations

• Josh Chamberlain add scraper philosophies to readme (do what you want, we can always refactor later)

Decisions

• Dolt is not our advanced analytics tool. Dolt is not our deep storage layer.