Replacing database names with variables in SSDT - visual-studio

Question

We are using SSDT to create a build and deploy pipeline using Git and ADO. In order for the solution to build, all cross-database references must have a corresponding Database Reference in the project and must be referred to using a [$(DatabaseVariable)]. This means that every time a change is made and synched using Schema Compare, the database names have to be replaced either manually or with a find-and-replace. There are of course numerous downsides to the find-and-replace approach, including the fact that it will find and replace references in the project files that should not be database variables. I'm hoping someone knows of a way to automate this process that doesn't involve such a brute force method as find-and-replace.
I have searched around extensively for this and found nothing helpful.
Here's an example view that contains cross-database references:
CREATE view [Migration].[vwCHILDS_Allegation_Allegation]
as
with src as (
select
cast('' as nvarchar(50)) CEAllegationIdentifier
, (select AllegationTypeId from [$(CWNS_Migration)].Allegation.AllegationType where Code = dfrvmi1.DestinationDataFieldReferenceValueCode) AllegationTypeId
, cast(1 as int) SourceSystemId
, cast(src.IDNBR as nvarchar(64)) SourceSystemIdentifier
, src.IDNBR SourceSystemIdentifier_Numeric
, case when src.CRET_DT_TM = '0001-01-01' then null else src.CRET_DT_TM end SourceSystemCreatedDateTime
, case when src.MOD_DT_TM = '0001-01-01' then null else src.MOD_DT_TM end SourceSystemModifiedDateTime
, (
select
max(pe.PersonId)
from
[$(CWNS_Migration)].PersonIdentity.PersonIdentifier pe
join [$(CHILDSDB2)].VLCHA.STAFF_USERID st on cast(st.FK_STAFFFK_PERSID as nvarchar(64)) = pe.Identifier
and pe.PersonIdentificationSystemId = 4
where
st.USERID = ltrim(rtrim(src.MOD_USR_ID))) SourceSystemModifiedPersonId
from
[$(CHILDSDB2)].VLCHA.ALGTN src
left join [$(DataCatalog)].dbo.DataFieldReferenceValueMappingInfo dfrvmi1 on dfrvmi1.SourceDataFieldReferenceValueDataFieldId = 216
and dfrvmi1.SourceDataFieldReferenceValueCode = ltrim(rtrim(src.FK_ALGTN_PRIORICTG))
and dfrvmi1.DestinationDataFieldReferenceValueDataFieldId = 20605
)
select
src.*
from
src
left join [$(CWNS_Migration)].Allegation.Allegation tgt on tgt.SourceSystemId = src.SourceSystemId and tgt.SourceSystemIdentifier = src.SourceSystemIdentifier
left join [$(CWNS_Migration)].Quarantine.Allegation q on q.SourceSystemId = src.SourceSystemId and q.SourceSystemIdentifier = src.SourceSystemIdentifier
and q.QExecutionId = 1
where
q.QExecutionId is null
and (
isnull(src.AllegationTypeId, 0) <> isnull(tgt.AllegationTypeId, 0)
or isnull(try_cast(src.SourceSystemCreatedDateTime as datetime), '') <> isnull(tgt.SourceSystemCreatedDateTime, '')
or isnull(try_cast(src.SourceSystemModifiedDateTime as datetime), '') <> isnull(tgt.SourceSystemModifiedDateTime, '')
or isnull(src.SourceSystemModifiedPersonId, 0) <> isnull(tgt.SourceSystemModifiedPersonId, 0)
)
I'm also hoping there's a way to avoid having Schema Compare always show the variables as differences with the database.

Answer 1

Ideally you need to include all databases as the reference to your project (right click on the project and do Add database reference, or something like that). You can do that in 2 ways:
* Create project for each database and import all or only objects that being used by your database
* Extract dacpac from the live database and use it as a reference
Then create synonyms for every single cross database objects in your main database. For example if we have object [$(CWNS_Migration)].PersonIdentity.PersonIdentifier then you'll need to create synonym as
CREATE SYNONYM PersonIdentity.PersonIdentifier
FOR [$(CWNS_Migration)].PersonIdentity.PersonIdentifier;
and then use PersonIdentity.PersonIdentifier (2 part name) in your code instead of the 3-4 part names. That's the project setup.
Now, when you right click on the project and do "Publish", it will pop up the dialogue where you can specify the vale for the CWNS_Migration variable and change it to whatever you need. After that you have the option to save this settings (Save As) and the result of that save is called as "Publish Profile". So, for every environment you just need to create different publish profile and use it when publish your changes.
You might want to try to start with synonyms first, it might work (need to verify), but I would recommend to add database references anyway.