Emap using MS Sql Server as a target (proof of concept) #90
Conversation
dialects. Also use docker compose profiles to turn containers on and off.
this is Postgres-specific.
…s SQL Server, but for H2 we need to make sure the column length is also specified in audit fields.
stefpiatek
left a comment
stefpiatek
left a comment
There was a problem hiding this comment.
Well that's fun, once we've got an example with this working let us know - would be keen to give to information services and have them be able to toy around with it
| InformDB: | ||
| emap: | ||
| branch: develop | ||
| hl7-vitals: |
There was a problem hiding this comment.
should this still be here?
There was a problem hiding this comment.
hl7-vitals? Not sure, I don't recognise that name. It might be trying to test having a different checkout directory to the repo name, but changing any of those values doesn't break any additional tests so who knows.
There was a problem hiding this comment.
this is an old name and should be removed
| ``` | ||
| - Timestamps (date and times) are timezone aware (and automated testing enforces this) and named in the form `<something>Datetime` | ||
| - Dates should only have date information and be named in the form `<something>Date` | ||
| - `timestamp with time zone` is Postgres-specific so should not be specified manually. Hibernate does the right thing in this case. |
SQL Server, use @lob annotation which maps to "text" and "varchar(max)" as appropriate.
stefpiatek
left a comment
stefpiatek
left a comment
There was a problem hiding this comment.
This is looking nice, Steve is really grateful to see this working 🙏🏻
| fakeuds-mssql: | ||
| image: mcr.microsoft.com/mssql/server:2019-CU32-ubuntu-20.04 | ||
| env_file: | ||
| - ../../config/fakeuds-mssql-config-envs | ||
| environment: | ||
| ACCEPT_EULA: "Y" | ||
| MSSQL_PID: "Express" | ||
| TZ: "Europe/London" | ||
| volumes: | ||
| - mssql-data-fakeuds:/var/opt/mssql | ||
| restart: unless-stopped | ||
| ports: | ||
| - "${FAKEUDS_MSSQL_PORT}:1433" | ||
| profiles: | ||
| - fakeuds-mssql | ||
|
|
There was a problem hiding this comment.
semantics but I'd argue that this isn't a fake UDS, instead its a mssql-star or just mssql-db
| # dialect to use for core informdb tables | ||
| UDS_HIBERNATE_DIALECT: uk.ac.ucl.rits.inform.informdb.PostgreSQLEmapDialect | ||
| # dialect for modules that don't have access to emap-star (which contains the above code) | ||
| UDS_HIBERNATE_DIALECT_2: org.hibernate.dialect.PostgreSQL95Dialect |
There was a problem hiding this comment.
Would this be a more meaningful name?
| UDS_HIBERNATE_DIALECT_2: org.hibernate.dialect.PostgreSQL95Dialect | |
| UDS_HIBERNATE_DIALECT_FALLBACK: org.hibernate.dialect.PostgreSQL95Dialect | |
| * ever needed to use more than one type of SQL array. Hopefully we will have | ||
| * found a better way of storing waveform data before then! | ||
| */ | ||
| public PostgreSQLEmapDialect() { |
easy to switch between them.
jeremyestein
force-pushed
the
jeremy/mssql
branch
from
cd01a52 to
b1d8ae9
Compare
|
(reason for force push: removing changes to deployment documentation that are now in their own PR #124 ) |
SQL Server did not like the following types we were manually specifying:
byteatimestamp with time zonedouble precision arrayThe general tactic has been to remove these manual types from
@Columnannotations, and create three custom SQL dialects: for postgres, SQL server, and H2 which specifies them instead on a per-backend basis.SQL Server does not in fact support SQL arrays, and I haven't attempted to reimplement them in another way, so waveform data will not work in SQL Server without further effort.
The current status is that the DDL runs nicely on my machine, but since we don't have any system tests (except for the waveform-generator which doesn't work in SQL Server anyway), I will have to test this on the GAE to see if real data gets put in.