Empty graph context variable

[mChlopek]

Hi there,
we are currently adapting sqlmesh with our dbt project.
Topic related to merged pr Feat(dbt): Add dbt graph context variable support #5159

I tried to use a graph variable in the sqlmesh plan but I got
empty values:

The same code run with the dbt run command returned values.
I was hoping it would work the same with sqlmesh plan.
Thanks in advance for help!

[themisvaltinos]

hey @mChlopek can you provide more information on what command you run in sqlmesh and the full macro and where this macro was invoked? Be aware that during parse time the graph is empty and it is populated and available at runtime. So was it maybe logged subsequently?

[mChlopek]

Hey themisvaltinos
I ran sqlmesh plan
In the SQL model file I have a cte where the code is created by the union_all_tagged_models macro.

I tried to reference the graph variable directly in union_all_tagged_models macro (empty)

I use graph variable to get all model objects and filter them by tag in get_models_with_tag_and_path

The final results is the SQL code that creates UNION ALL using all models with a specific tag

So I need this graph variable before generating the model code (based on historical dependencies?). Is there a way to handle this like it works in DBT?
Thank you for your help !

[themisvaltinos]

I see thank you so this pr : #5300 that is available on the latest v0.215.0 release https://github.com/TobikoData/sqlmesh/releases/tag/v0.215.0 should have resolved this issue. Let us know @mChlopek if upgrading to this worked for you

[themisvaltinos]

this pr also makes the graph available at parse time: #5329

[mChlopek]

Hi, themisvaltinos, it seems that after updating to version 0.215.0 (now 0.216) I still have an empty graph and I can't generate the union all of all tagged models :(
Maybe mentioned pr #5329 will handle this in next release, let you know

[themisvaltinos]

Maybe mentioned pr #5329 will handle this in next release, let you know

yes let us know if the next release which includes this does work

[mChlopek]

Hi themisvaltinos, it seems to work, thank you! You're rockstar!
I found another bug related to sqlmesh migration, but we are much closer to the end now
#5341

and others active issues:
#5295
#5328
#5262 merged, but it can't be found in the description of new releases