← Platform Fundamentals & Architecture Review

What changed: the 2025–2026 rename map

Databricks renamed at least eight products between June 2025 and July 2026. This lesson is your decoder ring. It's also the lesson with the shortest shelf life — see "Re-verifying" at the end.

The rename table

Reassuringly, the Declarative Automation Bundles rename was explicitly non-breaking: "the bundle CLI command and all of your existing configuration does not need to be modified."

The code change that actually breaks

Renames are cosmetic. This one isn't. The pipelines Python API changed:

# OLD — every pre-2025 DLT tutorial
import dlt

@dlt.table
def my_table():
    ...

@dlt.view
def my_view():
    ...

# CURRENT
from pyspark import pipelines as dp

@dp.table                       # streaming table — pairs with spark.readStream
def ingested():
    return spark.readStream.format("cloudFiles")...

@dp.materialized_view           # batch — pairs with spark.read
def rolled_up():
    return spark.read.table("ingested").groupBy(...)

@dp.temporary_view
def my_view():
    ...

The rule to memorize: spark.readStream + @dp.table → streaming table. spark.read + @dp.materialized_view → materialized view. The decorator and the read method must agree.

On import dlt, be precise — the docs say it is replaced and not recommended, but still functional, not removed:

"Previously, Databricks used the dlt module to support pipeline functionality. The dlt module has been replaced by the pyspark.pipelines module. You may still use dlt, but Databricks recommends using pipelines." — Pipelines Python reference (2026-07-27)

So if an exam question asks "has import dlt been removed?" the answer is no.

This aligns Databricks with Apache Spark Declarative Pipelines, which Databricks donated to the Apache Spark project and which lands in Spark 4.1. Source: Where is DLT? (updated 2026-07-10).

Retired and deprecated

• Community Edition — retired 2026-01-01. Any tutorial opening with "sign up for Community Edition" is pre-2026. The replacement is Free Edition.
• DBFS root and mounts — deprecated; new accounts are provisioned without access.
• Hive metastore — legacy. No published EOL date.

Coming deprecations to put in your calendar

From What's coming:

The freshness test

Before trusting any Databricks tutorial, blog, or video, scan for these. Each one dates the material:

1. Says "Delta Live Tables" or "DLT" → before mid-2025
2. Says import dlt → before mid-2025, and the code needs changing
3. Says "Community Edition" → before 2026
4. Says "Databricks Asset Bundles" → before March 2026
5. Says "Delta Sharing" → before June 2026
6. Says "Single user" / "Shared" cluster → older naming
7. Uses dbutils.fs.mount → won't run on a new account at all
8. Says "Databricks Assistant" → before March 2026

This test is worth more than any single fact in this course. Content ages; the ability to date content on sight doesn't.

An honest caveat about the docs

Two things worth knowing so you don't lose an afternoon:

Databricks retro-edits historical release notes. The June 2025 entry announcing the DLT rename today reads "Spark Declarative Pipelines on Lakeflow" — which is not what it said in June 2025. Treat any release-note quote as current wording, not historical wording.

The docs are currently inconsistent on pipeline naming. The AWS release-notes index says "Lakeflow Spark Declarative Pipelines," the GCP one says "Lakeflow pipelines," and the /ldp/ section is titled "Apache Spark™ Declarative Pipelines." All at the same time.

The pragmatic answer: say "Lakeflow pipelines, formerly DLT" and move on. Anyone who corrects you is quoting a different page of the same documentation set.

Re-verifying

This lesson decays fastest. Re-check monthly:

Swap /aws/ for /gcp/; Azure lives on learn.microsoft.com/en-us/azure/databricks/.

Check yourself

1. A video demonstrates "Delta Live Tables" with import dlt. Two problems — what are they?
2. Your team's runbook says "grant access via Delta Sharing." Is it wrong?
3. What breaks on 2026-08-22?

Teaching this section

• Teach this lesson FIRST if your audience is experienced. For people who used Databricks in 2023–2024, this is the highest-value 15 minutes in the module; everything else is revision.
• Best exercise: have learners search YouTube for "Databricks ETL tutorial," open the top result, and date it using the freshness test. It's usually stale, and finding that out themselves is the lesson.
• Tone: don't present churn as chaos. Frame it as why building your own continuously-updated material beats buying a fixed course — which is, after all, why this repo exists.
• Timing: 15 minutes presented, 10 for the dating exercise.