[Bug]: column block-list is changing in every plan (upgrade to v0.95.0 not possible

[JESCHO99]

Terraform CLI Version

1.9.1

Terraform Provider Version

v0.95.0

Terraform Configuration

terraform {
  required_version = "~> 1.9.0"
  required_providers {
    snowflake = {
      source  = "Snowflake-Labs/snowflake"
      version = "~> 0.95.0"
    }
  }
}

provider "snowflake" {
  role = "SYSADMIN"
}

# DATABASE
resource "snowflake_database" "this" {
  name = "TEST_DATABASE"
}

# SCHEMA
resource "snowflake_schema" "this" {
  database                    = snowflake_database.this.name
  name                        = "TEST_SCHEMA"
  is_transient                = "false"
  #with_managed_access         = "false"
  data_retention_time_in_days  = 1
}

# TABLES
resource "snowflake_table" "table_1" {
  database                    = snowflake_database.this.name
  schema                      = snowflake_schema.this.name
  name                        = "TEST_TABLE_V1"
  data_retention_time_in_days = 1
  change_tracking             = false

  column {
    name = "ID"
    type = "NUMBER(11,2)"
    nullable = false
  }

  column {
    name = "NAME"
    type = "VARCHAR(30)"
    nullable = true
  }
}

resource "snowflake_view" "this" {

  database = snowflake_database.this.name
  schema   = snowflake_schema.this.name
  name     = "TEST_VIEW"

  statement = <<-SQL
    SELECT ID, NAME FROM TEST_DATABASE.TEST_SCHEMA.TEST_TABLE_V1;
  SQL

  is_secure  = false

  depends_on = [
    snowflake_database.this,
    snowflake_schema.this,
    snowflake_table.table_1
  ]
}

Category

category:resource

Object type(s)

data_source:views

Expected Behavior

The following terraform configurations should be able to execute a terraform apply without any error and after a successful apply of this code there should not be any open changes left when a new terraform plan is executed after the successful apply

Actual Behavior

After a successful apply when a terraform plan is executed there is always a change for the view object for the column list. Even is the column list is not touched in the code this behaviour leads to an update/replacement of all views in every terraform plan executed.

This is the change/update which shows up after every successful apply in the next terraform plan when using this terraform version and view resources.

Steps to Reproduce

1. terraform apply
2. terraform plan
3. terraform apply
4. terraform plan

How much impact is this issue causing?

High

Logs

No response

Additional Information

We are not able to upgrade to version 0.95.0 because of this issue right now and for us it would be important because we need the hotfix for table data types which is implemented in the 0.95.0 release

Would you like to implement a fix?

• Yeah, I'll take it 😎

[sfc-gh-asawicki]

Hey @JESCHO99. Thanks for reaching out to us.

View resource is a newly reworked resource, so we will analyze this with a high priority. cc: @sfc-gh-jmichalak

[sfc-gh-jmichalak]

Hi @JESCHO99 👋,
We're missing a diff suppression for columns, so when the columns are determined by the referenced table, they are available in DESC and stored in the state. This will be fixed in the next release.

As a workaround, please specify columns in the config, like:

  column {
    column_name = "ID"
  }

  column {
    column_name = "NAME"
  }

[JESCHO99]

Hey @sfc-gh-jmichalak,
thanks for looking at this issue this fast. We have over 1000 views in our code base and putting all those columns into the config (for sure the code could be generated) means work for our team which we want/need to roll back when the problem is solved. Do you have an idea when the next release will be there so we can decide if we can wait so we do not need to implement the hotfix and then roll back the code changes?

Best regards,
Jens

[sfc-gh-jmichalak]

I think we can release the fix ~after Wednesday next week.

[liamjamesfoley]

Hitting the same issue, also appreciate the fast fix. Trying to stay current as the provider nears 1.0 🤞 Thanks!

[nx-rebecca]

Also encountering this issue - any update on the release date for the fix? @sfc-gh-jmichalak

[sfc-gh-jmichalak]

This will be released today or tomorrow.

[sfc-gh-jmichalak]

Hi all, we've released v0.96.0, which includes the fix and acceptance test for the column list in views. Please upgrade with the migration guide.

[gws]

Thanks for all your work on this, much appreciated.

v0.96.0 fixes the issue for me when no column blocks are specified, but I'm still hitting a plan loop when updating to v0.96.0 when specifying columns along with comments, for example, the following resource creates the correct view in Snowflake:

resource "snowflake_view" "example" {
  name = "EXAMPLE"

  database = snowflake_schema.this.database
  schema   = snowflake_schema.this.name

  comment   = snowflake_table.event_v0.comment
  ## The column identifiers in the SELECT list are unquoted
  statement = <<-SQL
    SELECT ${join(", ", [for c in local.view_columns_v0 : c.name])}
    FROM ${snowflake_table.source.name}
  SQL

  dynamic "column" {
    for_each = local.view_columns_v0

    content {
      column_name = column.value.name
      comment     = column.value.comment
    }
  }
}

... but on subsequent runs, I'm seeing same plan loop as before, with the perma-diff being on the show_output attribute of the view in the plan.

This seems to only be an issue when explicitly specifying the columns in the SELECT in statement, and also specifying a comment along with a column_name in the column blocks. For instance, a view using SELECT * appears to be unaffected.

This feels like the same issue with the v0.96.0 release containing a partial fix for the plan loop under some circumstances, but can file a new issue if not (or maybe I'm missing something obvious?)

[sfc-gh-jmichalak]

Hi @gws 👋

I tried to reproduce this, but I have no permadiff. Could you also provide the referenced table definition, the output of terraform plan, and logs with TF_LOG=DEBUG?

[gws]

@sfc-gh-jmichalak Thanks, this is a weirder one than I thought initially.

I think I can do you one better and get you something that reproduces the issue consistently for me on v0.96.0, I think the issue is related to () (with anything or nothing in between) appearing in the text of the comment on the view.

Not as minimal as it could be but should be self-contained enough to repro:

locals {
  db_name     = "TEST"
  schema_name = "GH3073"

  columns = [
    {
      comment  = "COL1 comment"
      name     = "COL1"
      nullable = false
      type     = "VARCHAR"
    },
    {
      ## This appears to be the source of the issue
      comment  = "COL2 comment has issues when the following (parentheses) appear."
      name     = "COL2"
      nullable = false
      type     = "VARCHAR"
    },
  ]

  view_columns = [for c in local.columns : c if c.name != "COL1"]
}

resource "snowflake_schema" "this" {
  database = local.db_name
  name     = local.schema_name
}

resource "snowflake_table" "test" {
  name = "T_TEST"

  database = snowflake_schema.this.database
  schema   = snowflake_schema.this.name

  dynamic "column" {
    for_each = local.columns

    content {
      comment  = column.value.comment
      name     = column.value.name
      nullable = column.value.nullable
      type     = column.value.type
    }
  }
}

resource "snowflake_view" "test" {
  name = "V_TEST"

  database = snowflake_schema.this.database
  schema   = snowflake_schema.this.name

  statement = <<-SQL
    SELECT ${join(", ", [for c in local.view_columns : c.name])}
    FROM ${snowflake_table.test.name}
  SQL

  dynamic "column" {
    for_each = local.view_columns

    content {
      column_name = column.value.name
      comment     = column.value.comment
    }
  }
}

Terraform plan output:

Terraform will perform the following actions:

  # module.gh3073.snowflake_view.test will be updated in-place
  ~ resource "snowflake_view" "test" {
        id                   = "\"TEST\".\"GH3073\".\"V_TEST\""
        name                 = "V_TEST"
      ~ show_output          = [
          - {
              - change_tracking = "OFF"
              - comment         = ""
              - created_on      = "2024-09-20T22:32:09.199-07:00"
              - database_name   = "TEST"
              - is_materialized = false
              - is_secure       = false
              - kind            = ""
              - name            = "V_TEST"
              - owner           = "SYSADMIN"
              - owner_role_type = "ROLE"
              - reserved        = ""
              - schema_name     = "GH3073"
              - text            = <<-EOT
                    CREATE OR REPLACE VIEW "TEST"."GH3073"."V_TEST" ("COL2" COMMENT 'COL2 comment has issues when the following (parentheses) appear.') AS SELECT COL2
                    FROM T_TEST
                EOT
            },
        ] -> (known after apply)
      ~ statement            = <<-EOT
          - appear.') AS SELECT COL2
          + SELECT COL2
            FROM T_TEST
        EOT
        # (9 unchanged attributes hidden)

        # (1 unchanged block hidden)
    }

Plan: 0 to add, 1 to change, 0 to destroy.

If you still can't reproduce with the above I'll get the logs over to you as well but I suspect we won't need them and they're pretty verbose.

[sfc-gh-jmichalak]

Thanks! We need to parse manually the text field to get view statement. This seems to be an issue with the parser. As a workaround, please don't use () in the comment until this is fixed.

[liamjamesfoley]

We're seeing lots of SQL compilation error: Invalid column definition list errors now on 0.96 because the provider doesn't seem to be updating the column definition list correctly. Is that related to this issue or separate?

For example, the PR screenshoted below added a column to a view called "SNPS_GROUP_NUM" but it doesn't appear in the definition list so the query fails.

[sfc-gh-jmichalak]

Please make sure you have a matching column list with the columns returned from the statement. I encountered many errors like this during testing, and usually, the problem was bad configuration. If it doesn't work, please provide the whole resource configuration.

[liamjamesfoley]

Hi @sfc-gh-jmichalak, we aren't setting the column block list at all.

We have a hundreds of views that use a module that makes it so people w/ little to no terraform experience can easily manage views by just managing one .SQL file.

The module takes sql files in a directory like database_name/schema_name/view_name.sql
and generates the database name, the schema name, and the view name from the file path and uses the contents as the definition, and parses out comments from the top of the file

resource "snowflake_view" "views" {
  for_each = {
    for path in fileset(var.base_dir, "**/*.sql") : path => {
      database = upper(split("/", path)[0])
      schema   = upper(split("/", path)[1])
      name     = upper(trimsuffix(split("/", path)[2], ".sql"))
      comment  = format("Managed by Terraform: %s", try(trimspace(regex("//(.*)", file("${var.base_dir}/${path}"))[0]), ""))
    }
  }
  database  = each.value.database
  schema    = each.value.schema
  name      = each.value.name
  comment   = each.value.comment
  statement = file("${var.base_dir}/${each.key}")
  # or_replace = true
  is_secure       = "false"
  is_recursive    = "false"
  is_temporary    = "false"
  change_tracking = "false"
}

---

variable "base_dir" {
  type = string
}

[sfc-gh-jmichalak]

Hi, I just tested your configuration with a simple SELECT id FROM "DB"."SCHEMA"."TEST_2";, and it works on my side. My suggestion is to

• paste the logs with TF_LOG=DEBUG
• check if the view statement is correct - specifically if the columns in the statement match the underlying tables. You can check if executing the statement alone (SELECT ...) in Snowsight returns no errors.

When the column definition is empty in the config (as in your case), the column list in CREATE should be omitted and inferred on the Snowflake side from the statement.

[liamjamesfoley]

@sfc-gh-jmichalak Thanks, will do if it comes up again!

[liamjamesfoley]

Hey @sfc-gh-jmichalak this just happened again.

Terraform changed the actual definition but not the column_list, which is a very confusing experience for the end users.

Here's the change:

Here's the TF Plan:

Terraform used the selected providers to generate the following execution
plan. Resource actions are indicated with the following symbols:
  ~ update in-place

Terraform will perform the following actions:

  # module.views.snowflake_view.views["analytics/omx/member_status_history_view.sql"] will be updated in-place
  ~ resource "snowflake_view" "views" {
        id                   = "\"ANALYTICS\".\"OMX\".\"MEMBER_STATUS_HISTORY_VIEW\""
        name                 = "MEMBER_STATUS_HISTORY_VIEW"
      ~ show_output          = [
          - {
              - change_tracking = "OFF"
              - comment         = "Managed by Terraform: View for Materializing member_status_history"
              - created_on      = "2024-09-19T17:19:03.057Z"
              - database_name   = "ANALYTICS"
              - is_materialized = false
              - is_secure       = false
              - kind            = ""
              - name            = "MEMBER_STATUS_HISTORY_VIEW"
              - owner           = "ANALYTICS__OMX__WRITER"
              - owner_role_type = "ROLE"
              - reserved        = ""
              - schema_name     = "OMX"
              - text            = <<-EOT
                    CREATE OR REPLACE VIEW "ANALYTICS"."OMX"."MEMBER_STATUS_HISTORY_VIEW" ("ticket_id", "member_id", "CREATED_AT", "DATE_KEY", "LAST_DD_DT", "LAST_DD_SOURCE", "Member Status at interaction", "Member cancellation reason at interaction", "Member suspension reason at interaction", "Limited access tier", "Step up auth") COPY GRANTS comment = 'Managed by Terraform: View for Materializing member_status_history' AS //View for Materializing member_status_history
                    SELECT distinct
                              zendesk_ticket."ID"::string  AS "ticket_id",
                              zendesk_ticket.member_id  AS "member_id",
                              zendesk_ticket."CREATED_AT" ,
                              member.DATE_KEY,
                              member.LAST_DD_DT,
                              member.LAST_DD_SOURCE,
                              coalesce(member.user_status,member2.status) AS "Member Status at interaction",
                              coalesce(member.cancellation_reason, member2.cancellation_reason) as "Member cancellation reason at interaction",
                              coalesce(member.suspension_reason, member2.suspension_reason) AS "Member suspension reason at interaction",
                              iff(lat.user_id is not null , TRUE, false) as "Limited access tier",
                              iff(SUA.user_id is not null , TRUE, false) as "Step up auth"
                    
                        FROM analytics.looker.zendesk_tickets_base  AS zendesk_ticket
                        LEFT JOIN edw_db.core.member_details_history AS member ON zendesk_ticket.member_id= member.user_id AND replace(TO_CHAR(TO_DATE(zendesk_ticket."CREATED_AT" ), 'YYYY-MM-DD'),'-')=date_key
                        LEFT JOIN chime.finance.members AS member2 ON zendesk_ticket.member_id=member2.id AND zendesk_ticket.created_at>=member2.updated_at
                        left join analytics.omx.limited_access_tier as LAT on zendesk_ticket.member_id = lat.user_id and zendesk_ticket.created_at between LAT.lat_added_timestamp and LAT.lat_removed_timestamp
                        left join analytics.omx.step_up_auth as SUA on zendesk_ticket.member_id = SUA.user_id and zendesk_ticket.created_at between SUA.sua_status_created_at and coalesce(sua.login_success_timestamp,sua.canceled_by_agent_timestamp)
                          ;
                EOT
            },
        ] -> (known after apply)
      ~ statement            = <<-EOT
            //View for Materializing member_status_history
            SELECT distinct
          -           zendesk_ticket."ID"::string  AS "ticket_id",
          -           zendesk_ticket.member_id  AS "member_id",
          -           zendesk_ticket."CREATED_AT" ,
          +           zendesk_ticket."ID"::string  AS ticket_id,
          +           zendesk_ticket.member_id  AS member_id,
          +           zendesk_ticket.CREATED_AT ,
                      member.DATE_KEY,
                      member.LAST_DD_DT,
                      member.LAST_DD_SOURCE,
          -           coalesce(member.user_status,member2.status) AS "Member Status at interaction",
          -           coalesce(member.cancellation_reason, member2.cancellation_reason) as "Member cancellation reason at interaction",
          -           coalesce(member.suspension_reason, member2.suspension_reason) AS "Member suspension reason at interaction",
          -           iff(lat.user_id is not null , TRUE, false) as "Limited access tier",
          -           iff(SUA.user_id is not null , TRUE, false) as "Step up auth"
          - 
          +           coalesce(member.user_status,member2.status) AS member_status_at_interaction,
          +           coalesce(member.cancellation_reason, member2.cancellation_reason) as member_cancellation_reason_at_interaction,
          +           coalesce(member.suspension_reason, member2.suspension_reason) AS member_suspension_reason_at_interaction,
          +           iff(lat.user_id is not null , TRUE, false) as limited_access_tier,
          +           iff(SUA.user_id is not null , TRUE, false) as step_up_auth
                FROM analytics.looker.zendesk_tickets_base  AS zendesk_ticket
                LEFT JOIN edw_db.core.member_details_history AS member ON zendesk_ticket.member_id= member.user_id AND replace(TO_CHAR(TO_DATE(zendesk_ticket."CREATED_AT" ), 'YYYY-MM-DD'),'-')=date_key
                LEFT JOIN chime.finance.members AS member2 ON zendesk_ticket.member_id=member2.id AND zendesk_ticket.created_at>=member2.updated_at
                left join analytics.omx.limited_access_tier as LAT on zendesk_ticket.member_id = lat.user_id and zendesk_ticket.created_at between LAT.lat_added_timestamp and LAT.lat_removed_timestamp
                left join analytics.omx.step_up_auth as SUA on zendesk_ticket.member_id = SUA.user_id and zendesk_ticket.created_at between SUA.sua_status_created_at and coalesce(sua.login_success_timestamp,sua.canceled_by_agent_timestamp)
          -       ;
        EOT
        # (10 unchanged attributes hidden)

        # (11 unchanged blocks hidden)
    }

Plan: 0 to add, 1 to change, 0 to destroy.

Here's the Query generated by Terraform:

CREATE OR REPLACE VIEW "ANALYTICS"."OMX"."MEMBER_STATUS_HISTORY_VIEW" ("ticket_id", "member_id", "CREATED_AT", "DATE_KEY", "LAST_DD_DT", "LAST_DD_SOURCE", "Member Status at interaction", "Member cancellation reason at interaction", "Member suspension reason at interaction", "Limited access tier", "Step up auth") COPY GRANTS COMMENT = 'Managed by Terraform: View for Materializing member_status_history' AS //View for Materializing member_status_history
SELECT distinct
          zendesk_ticket."ID"::string  AS ticket_id,
          zendesk_ticket.member_id  AS member_id,
          zendesk_ticket.CREATED_AT ,
          member.DATE_KEY,
          member.LAST_DD_DT,
          member.LAST_DD_SOURCE,
          coalesce(member.user_status,member2.status) AS member_status_at_interaction,
          coalesce(member.cancellation_reason, member2.cancellation_reason) as member_cancellation_reason_at_interaction,
          coalesce(member.suspension_reason, member2.suspension_reason) AS member_suspension_reason_at_interaction,
          iff(lat.user_id is not null , TRUE, false) as limited_access_tier,
          iff(SUA.user_id is not null , TRUE, false) as step_up_auth
    FROM analytics.looker.zendesk_tickets_base  AS zendesk_ticket
    LEFT JOIN edw_db.core.member_details_history AS member ON zendesk_ticket.member_id= member.user_id AND replace(TO_CHAR(TO_DATE(zendesk_ticket."CREATED_AT" ), 'YYYY-MM-DD'),'-')=date_key
    LEFT JOIN chime.finance.members AS member2 ON zendesk_ticket.member_id=member2.id AND zendesk_ticket.created_at>=member2.updated_at
    left join analytics.omx.limited_access_tier as LAT on zendesk_ticket.member_id = lat.user_id and zendesk_ticket.created_at between LAT.lat_added_timestamp and LAT.lat_removed_timestamp
    left join analytics.omx.step_up_auth as SUA on zendesk_ticket.member_id = SUA.user_id and zendesk_ticket.created_at between SUA.sua_status_created_at and coalesce(sua.login_success_timestamp,sua.canceled_by_agent_timestamp)

Heres' the current definition after the TF apply.

CREATE OR REPLACE VIEW "ANALYTICS"."OMX"."MEMBER_STATUS_HISTORY_VIEW" ("ticket_id", "member_id", "CREATED_AT", "DATE_KEY", "LAST_DD_DT", "LAST_DD_SOURCE", "Member Status at interaction", "Member cancellation reason at interaction", "Member suspension reason at interaction", "Limited access tier", "Step up auth") COPY GRANTS comment = 'Managed by Terraform: View for Materializing member_status_history' AS //View for Materializing member_status_history
SELECT distinct
          zendesk_ticket."ID"::string  AS ticket_id,
          zendesk_ticket.member_id  AS member_id,
          zendesk_ticket.CREATED_AT ,
          member.DATE_KEY,
          member.LAST_DD_DT,
          member.LAST_DD_SOURCE,
          coalesce(member.user_status,member2.status) AS member_status_at_interaction,
          coalesce(member.cancellation_reason, member2.cancellation_reason) as member_cancellation_reason_at_interaction,
          coalesce(member.suspension_reason, member2.suspension_reason) AS member_suspension_reason_at_interaction,
          iff(lat.user_id is not null , TRUE, false) as limited_access_tier,
          iff(SUA.user_id is not null , TRUE, false) as step_up_auth
    FROM analytics.looker.zendesk_tickets_base  AS zendesk_ticket
    LEFT JOIN edw_db.core.member_details_history AS member ON zendesk_ticket.member_id= member.user_id AND replace(TO_CHAR(TO_DATE(zendesk_ticket."CREATED_AT" ), 'YYYY-MM-DD'),'-')=date_key
    LEFT JOIN chime.finance.members AS member2 ON zendesk_ticket.member_id=member2.id AND zendesk_ticket.created_at>=member2.updated_at
    left join analytics.omx.limited_access_tier as LAT on zendesk_ticket.member_id = lat.user_id and zendesk_ticket.created_at between LAT.lat_added_timestamp and LAT.lat_removed_timestamp
    left join analytics.omx.step_up_auth as SUA on zendesk_ticket.member_id = SUA.user_id and zendesk_ticket.created_at between SUA.sua_status_created_at and coalesce(sua.login_success_timestamp,sua.canceled_by_agent_timestamp)

Here's the Describe statement after the tf-apply:

[liamjamesfoley]

Here's another issue, but this time the apply fails.

Change:

Plan:

Terraform used the selected providers to generate the following execution
plan. Resource actions are indicated with the following symbols:
  ~ update in-place

Terraform will perform the following actions:

  # module.views.snowflake_view.views["hightouch_db/chime_data/conversion_signals_v2_view.sql"] will be updated in-place
  ~ resource "snowflake_view" "views" {
        id                   = "\"HIGHTOUCH_DB\".\"CHIME_DATA\".\"CONVERSION_SIGNALS_V2_VIEW\""
        name                 = "CONVERSION_SIGNALS_V2_VIEW"
      ~ show_output          = [
          - {
              - change_tracking = "OFF"
              - comment         = "Managed by Terraform: "
              - created_on      = "2024-09-23T20:42:22.804Z"
              - database_name   = "HIGHTOUCH_DB"
              - is_materialized = false
              - is_secure       = false
              - kind            = ""
              - name            = "CONVERSION_SIGNALS_V2_VIEW"
              - owner           = "HIGHTOUCH_DB__CHIME_DATA__WRITER"
              - owner_role_type = "ROLE"
              - reserved        = ""
              - schema_name     = "CHIME_DATA"
              - text            = <<-EOT
                    CREATE OR REPLACE VIEW "HIGHTOUCH_DB"."CHIME_DATA"."CONVERSION_SIGNALS_V2_VIEW" ("COMP_KEY", "USER_ID", "DW_PROCESSING_TIMESTAMP", "GCLID", "WBRAID", "GBRAID", "EM", "PH", "FBC", "FBP", "CONTEXT_IP", "CONTEXT_USER_AGENT", "FN", "LN", "EXTERNAL_ID", "EVENT_ID", "CONVERSION_TIME", "CONVERSION_TYPE", "RAW_CONVERSION_VALUE", "CONVERSION_VALUE") comment = 'Managed by Terraform: ' AS select 
                        comp_key,
                        user_id,
                        dw_processing_timestamp,
                        gclid,
                        wbraid,
                        gbraid,
                        em,
                        ph,
                        fbc,
                        fbp,
                        context_ip,
                        context_user_agent,
                        fn,
                        ln,
                        external_id,
                        event_id,
                        conversion_time,
                        conversion_type,
                        raw_conversion_value,
                        conversion_value
                    from edw_db.marketing.conversion_signals_v2
                EOT
            },
        ] -> (known after apply)
      ~ statement            = <<-EOT
            select 
                comp_key,
                user_id,
                dw_processing_timestamp,
                gclid,
                wbraid,
                gbraid,
          +     click_id,
                em,
                ph,
                fbc,
                fbp,
                context_ip,
                context_user_agent,
                fn,
                ln,
                external_id,
                event_id,
                conversion_time,
                conversion_type,
                raw_conversion_value,
                conversion_value
            from edw_db.marketing.conversion_signals_v2
        EOT
        # (10 unchanged attributes hidden)

        # (20 unchanged blocks hidden)
    }

Plan: 0 to add, 1 to change, 0 to destroy.

apply:

Terraform v1.7.1
on linux_amd64
Executing pre-apply hook...
Enabling ECS task protection for this agent
{"protection":{"ExpirationDate":"2024-10-04T00:11:40.867Z","ProtectionEnabled":true,"TaskArn":"arn:aws:ecs:us-east-1:802476504392:task/d51c56121eaa44fc859986e22b373605"}}module.views.snowflake_view.views["hightouch_db/chime_data/conversion_signals_v2_view.sql"]: Modifying... [id="HIGHTOUCH_DB"."CHIME_DATA"."CONVERSION_SIGNALS_V2_VIEW"]
╷
│ Error: error creating view CONVERSION_SIGNALS_V2_VIEW err = 002026 (42601): SQL compilation error:
│ Invalid column definition list
│ 
│   with module.views.snowflake_view.views["hightouch_db/chime_data/conversion_signals_v2_view.sql"],
│   on ../../stack/snowflake/views/main.tf line 1, in resource "snowflake_view" "views":
│    1: resource "snowflake_view" "views" {
│ 
╵
Executing post-apply hook...
Disabling ECS task protection for this agent
{"protection":{"ExpirationDate":null,"ProtectionEnabled":false,"TaskArn":"arn:aws:ecs:us-east-1:802476504392:task/d51c56121eaa44fc859986e22b373605"}}Operation failed: failed running terraform apply (exit 1)

and the column in question, click_id, definitely exists in the source table.

[sfc-gh-jmichalak]

Hi @liamjamesfoley

Thanks for the detailed logs. This is indeed an error on our side (both of the cases). It happens because the old column list is still populated during CREATE OR REPLACE, which gets called when the statement is changed. This causes incorrect column definition in this SQL statement. This column list should be empty because it's not specified in the config.

The workaround here is to manually define the column list in the Terraform so that the names in the config match the names in the statement (please take care of proper quoting: in the statement, you use unquoted columns, so I believe in the column config they should be upper case). The other way is to get rid of the old state: remove the resource from the state and import it back. This should ensure that the columns in the state are correct.

Currently, I'm working on fixing this behavior.

[liamjamesfoley]

@sfc-gh-jmichalak Awesome, thanks for the confirmation! Do you think the fix will be part of 0.97?