diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml
diff --git a/.github/workflows/test-report.yml b/.github/workflows/test-report.yml
@@ -0,0 +1,24 @@
+name: Test Report
+on:
+  workflow_run:
+    workflows: ['CI']
+    types:
+      - completed
+jobs:
+  report:
+    if: ${{ github.event.workflow_run.conclusion == 'success' || github.event.workflow_run.conclusion == 'failure' }}
+    runs-on: ubuntu-latest
+    concurrency: report
+    steps:
+      - name: Download artifact
+        uses: dawidd6/action-download-artifact@v2
+        with:
+          workflow: ${{ github.event.workflow_run.workflow_id }}
+          workflow_conclusion: completed
+          path: artifacts
+
+      - name: publish test report
+        uses: EnricoMi/publish-unit-test-result-action@v1
+        with:
+          commit: ${{ github.event.workflow_run.head_sha }}
+          files: artifacts/**/junit.xml
diff --git a/.gitignore b/.gitignore
@@ -67,3 +67,6 @@ ibis_testing*
 .ipynb_checkpoints/
 .pytest_cache
 .mypy_cache
+
+# temporary doc build
+docbuild
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -1,16 +1,24 @@
 repos:
-  - repo: https://github.com/timothycrosley/isort
-    rev: 5.6.4
+  - repo: https://github.com/PyCQA/isort
+    rev: 5.9.3
     hooks:
       - id: isort
   - repo: https://github.com/psf/black
-    rev: 19.10b0
+    rev: 21.9b0
     hooks:
       - id: black
-        exclude: (ibis/_version|versioneer).py
+        exclude: (ibis/_version|versioneer)\.py
   - repo: git://github.com/pre-commit/pre-commit-hooks
     rev: v2.1.0
     hooks:
       - id: flake8
         types:
           - python
+  - repo: https://github.com/asottile/pyupgrade
+    rev: v2.29.0
+    hooks:
+      - id: pyupgrade
+        args: [--py37-plus]
+        exclude: (ibis/_version|versioneer)\.py
+        types:
+          - python
diff --git a/Makefile b/Makefile
diff --git a/README.md b/README.md
@@ -5,7 +5,7 @@
 | Documentation  | [![Documentation Status](https://img.shields.io/badge/docs-docs.ibis--project.org-blue.svg)](http://ibis-project.org) |
 | Conda packages | [![Anaconda-Server Badge](https://anaconda.org/conda-forge/ibis-framework/badges/version.svg)](https://anaconda.org/conda-forge/ibis-framework) |
 | PyPI           | [![PyPI](https://img.shields.io/pypi/v/ibis-framework.svg)](https://pypi.org/project/ibis-framework) |
-| Azure          | [![Azure Status](https://dev.azure.com/ibis-project/ibis/_apis/build/status/ibis-project.ibis)](https://dev.azure.com/ibis-project/ibis/_build) |
+| GitHub Actions | [![Build status](https://github.com/ibis-project/ibis/actions/workflows/main.yml/badge.svg)](https://github.com/ibis-project/ibis/actions/workflows/main.yml) |
 | Coverage       | [![Codecov branch](https://img.shields.io/codecov/c/github/ibis-project/ibis/master.svg)](https://codecov.io/gh/ibis-project/ibis) |
 
 
@@ -32,12 +32,13 @@ Ibis currently provides tools for interacting with the following systems:
 - [Apache Kudu](https://kudu.apache.org/)
 - [Hadoop Distributed File System (HDFS)](https://hadoop.apache.org/)
 - [PostgreSQL](https://www.postgresql.org/)
-- [MySQL](https://www.mysql.com/) (Experimental)
+- [MySQL](https://www.mysql.com/)
 - [SQLite](https://www.sqlite.org/)
 - [Pandas](https://pandas.pydata.org/) [DataFrames](http://pandas.pydata.org/pandas-docs/stable/dsintro.html#dataframe)
 - [Clickhouse](https://clickhouse.yandex)
 - [BigQuery](https://cloud.google.com/bigquery)
 - [OmniSciDB](https://www.omnisci.com)
-- [Spark](https://spark.apache.org) (Experimental)
+- [PySpark](https://spark.apache.org)
+- [Dask](https://dask.org/) (Experimental)
 
 Learn more about using the library at http://ibis-project.org.
diff --git a/azure-pipelines.yml b/azure-pipelines.yml
diff --git a/benchmarks/benchmarks.py b/benchmarks/benchmarks.py
@@ -138,7 +138,7 @@ def time_impala_large_expr_compile(self):
 
 class PandasBackend:
     def setup(self):
-        n = 30 * int(2e5)
+        n = 30 * int(2e4)
         self.data = pd.DataFrame(
             {
                 'key': np.random.choice(16000, size=n),
@@ -268,10 +268,10 @@ def time_high_card_grouped_rolling(self):
         self.high_card_grouped_rolling.execute()
 
     def time_low_card_grouped_rolling_udf(self):
-        self.low_card_grouped_rolling_udf.execute()
+        self.low_card_grouped_rolling_udf_mean.execute()
 
     def time_high_card_grouped_rolling_udf(self):
-        self.high_card_grouped_rolling_udf.execute()
+        self.high_card_grouped_rolling_udf_mean.execute()
 
     def time_low_card_window_analytics_udf(self):
         self.low_card_window_analytics_udf.execute()

diff --git a/ci/.env b/ci/.env
@@ -18,8 +18,3 @@ IBIS_TEST_POSTGRES_DATABASE=ibis_testing
 IBIS_TEST_CLICKHOUSE_HOST=clickhouse
 IBIS_TEST_CLICKHOUSE_PORT=9000
 IBIS_TEST_CLICKHOUSE_DATABASE=ibis_testing
-IBIS_TEST_OMNISCIDB_HOST=omniscidb
-IBIS_TEST_OMNISCIDB_PORT=6274
-IBIS_TEST_OMNISCIDB_DATABASE=ibis_testing
-IBIS_TEST_OMNISCIDB_USER=admin
-IBIS_TEST_OMNISCIDB_PASSWORD=HyperInteractive
diff --git a/ci/Dockerfile.dev b/ci/Dockerfile.dev
diff --git a/ci/Dockerfile.docs b/ci/Dockerfile.docs
diff --git a/ci/azure/linux.yml b/ci/azure/linux.yml
diff --git a/ci/backends-markers.sh b/ci/backends-markers.sh
diff --git a/ci/backends-to-start.sh b/ci/backends-to-start.sh
diff --git a/ci/check-services.sh b/ci/check-services.sh
diff --git a/ci/condarc b/ci/condarc
@@ -0,0 +1,24 @@
+# vim: ft=yaml
+always_yes: true
+
+# remote_connect_timeout_secs (float)
+#   The number seconds conda will wait for your client to establish a
+#   connection to a remote url resource.
+remote_connect_timeout_secs: 30.0
+
+# remote_max_retries (int)
+#   The maximum number of retries each HTTP connection should attempt.
+#
+remote_max_retries: 10
+
+# remote_backoff_factor (int)
+#   The factor determines the time HTTP connection should wait for
+#   attempt.
+#
+remote_backoff_factor: 2
+
+# remote_read_timeout_secs (float)
+#   Once conda has connected to a remote resource and sent an HTTP
+#   request, the read timeout is the number of seconds conda will wait for
+#   the server to send a response.
+remote_read_timeout_secs: 60.0
diff --git a/ci/datamgr.py b/ci/datamgr.py
diff --git a/ci/deps/bigquery.yml b/ci/deps/bigquery.yml
diff --git a/ci/deps/clickhouse.yml b/ci/deps/clickhouse.yml
@@ -1,5 +1,7 @@
-sqlalchemy>=1.3
-clickhouse-cityhash
-clickhouse-driver>=0.1.3
-clickhouse-sqlalchemy
-lz4
+name: ibis
+dependencies:
+  - sqlalchemy=1.3
+  - clickhouse-cityhash
+  - clickhouse-driver
+  - clickhouse-sqlalchemy
+  - lz4
diff --git a/ci/deps/dask-min.yml b/ci/deps/dask-min.yml
@@ -0,0 +1,4 @@
+name: ibis
+dependencies:
+  - dask=2021.2.0
+  - pyarrow
diff --git a/ci/deps/dask.yml b/ci/deps/dask.yml
@@ -0,0 +1,6 @@
+# For exclusions see https://github.com/ibis-project/ibis/pull/2802
+# dask 2021.5.0,5.1,6.0 had some issues with meta
+name: ibis
+dependencies:
+  - dask!=2021.5.*,!=2021.6.0
+  - pyarrow
diff --git a/ci/deps/hdf5.yml b/ci/deps/hdf5.yml
@@ -0,0 +1,3 @@
+name: ibis
+dependencies:
+  - pytables
diff --git a/ci/deps/impala.yml b/ci/deps/impala.yml
@@ -1,6 +1,8 @@
-sqlalchemy>=1.3
-impyla>=0.15.0
-requests>=2.24
-thrift>=0.9.3
-thriftpy2>=0.4
-thrift_sasl>=0.2.1
+name: ibis
+dependencies:
+  - bitarray=2.0.1
+  - impyla
+  - requests
+  - thrift_sasl
+  - python-hdfs
+  - boost
diff --git a/ci/deps/mysql.yml b/ci/deps/mysql.yml
@@ -1,2 +1,4 @@
-sqlalchemy>=1.3
-pymysql
+name: ibis
+dependencies:
+  - sqlalchemy
+  - pymysql
diff --git a/ci/deps/omniscidb.yml b/ci/deps/omniscidb.yml
diff --git a/ci/deps/parquet.yml b/ci/deps/parquet.yml
@@ -1 +1,3 @@
-pyarrow>=0.13
+name: ibis
+dependencies:
+  - pyarrow
diff --git a/ci/deps/postgres-min.yml b/ci/deps/postgres-min.yml
@@ -0,0 +1,7 @@
+name: ibis
+dependencies:
+  - sqlalchemy=1.3
+  - psycopg2=2.7
+  - geoalchemy2=0.6
+  - geopandas=0.6
+  - shapely=1.6
diff --git a/ci/deps/postgres.yml b/ci/deps/postgres.yml
@@ -1,3 +1,7 @@
-sqlalchemy>=1.3
-psycopg2>=2.8
-geoalchemy2>=0.6
+name: ibis
+dependencies:
+  - sqlalchemy
+  - psycopg2
+  - geoalchemy2
+  - geopandas
+  - shapely
diff --git a/ci/deps/pyspark-min.yml b/ci/deps/pyspark-min.yml
@@ -1,4 +1,6 @@
-# Need to import double-conversion below otherwise `import pyarrow` fails with ImportError: libdouble-conversion.so.3
-double-conversion
-pyarrow=0.12.1
-pyspark=2.4.3
+name: ibis
+dependencies:
+  - openjdk=8
+  - pyarrow=1
+  - pyspark=2.4.3
+  - pandas=1.2.5
diff --git a/ci/deps/pyspark.yml b/ci/deps/pyspark.yml
@@ -1 +1,3 @@
-pyspark>=2.4.3
+name: ibis
+dependencies:
+  - pyspark
diff --git a/ci/deps/spark-min.yml b/ci/deps/spark-min.yml
diff --git a/ci/deps/spark.yml b/ci/deps/spark.yml
diff --git a/ci/docker-compose.yml b/ci/docker-compose.yml
diff --git a/ci/dockerize.sh b/ci/dockerize.sh
diff --git a/ci/docs.sh b/ci/docs.sh
diff --git a/ci/impalamgr.py b/ci/impalamgr.py
@@ -30,13 +30,13 @@
 
 env_items = ENV.items()
 maxlen = max(map(len, map(toolz.first, env_items))) + len('IbisTestEnv[""]')
-format_string = '%-{:d}s == %r'.format(maxlen)
+format_string = f'%-{maxlen:d}s == %r'
 for key, value in env_items:
-    logger.info(format_string, 'IbisTestEnv[{!r}]'.format(key), value)
+    logger.info(format_string, f'IbisTestEnv[{key!r}]', value)
 
 
 def make_ibis_client(env):
-    hc = ibis.hdfs_connect(
+    hc = ibis.impala.hdfs_connect(
         host=env.nn_host,
         port=env.webhdfs_port,
         auth_mechanism=env.auth_mechanism,
@@ -55,16 +55,11 @@ def make_ibis_client(env):
     )
 
 
-def can_write_to_hdfs(con):
+def raise_if_cannot_write_to_hdfs(con):
     test_path = os.path.join(ENV.test_data_dir, ibis.util.guid())
     test_file = BytesIO(ibis.util.guid().encode('utf-8'))
-    try:
-        con.hdfs.put(test_path, test_file)
-        con.hdfs.rm(test_path)
-        return True
-    except Exception:
-        logger.exception('Could not write to HDFS')
-        return False
+    con.hdfs.put(test_path, test_file)
+    con.hdfs.rm(test_path)
 
 
 def can_build_udfs():
@@ -232,7 +227,7 @@ def upload_udfs(con):
 # ==========================================
 
 
-@click.group(context_settings=dict(help_option_names=['-h', '--help']))
+@click.group(context_settings={'help_option_names': ['-h', '--help']})
 def main():
     """Manage impala test data for Ibis."""
 
@@ -260,8 +255,8 @@ def load(data, udf, data_dir, overwrite):
     con = make_ibis_client(ENV)
 
     # validate our environment before performing possibly expensive operations
-    if not can_write_to_hdfs(con):
-        raise IbisError('Failed to write to HDFS; check your settings')
+    raise_if_cannot_write_to_hdfs(con)
+
     if udf and not can_build_udfs():
         raise IbisError('Build environment does not support building UDFs')
 

diff --git a/ci/load-data.sh b/ci/load-data.sh
diff --git a/ci/merge_and_update_env.sh b/ci/merge_and_update_env.sh
@@ -0,0 +1,28 @@
+#!/usr/bin/env bash
+#
+# Merge environment files and update the corresponding conda environment
+
+set -euo pipefail
+
+if [ "$#" -eq 0 ]; then
+  >&2 echo "error: must provide at least one backend"
+  exit 1
+fi
+
+# install conda-merge, don't try to update already installed dependencies
+mamba install --freeze-installed --name ibis conda-merge
+
+additional_env_files=()
+
+# pull all files associated with input backends
+for backend in "$@"; do
+  env_file="ci/deps/${backend}.yml"
+
+  if [ -f "${env_file}" ]; then
+    additional_env_files+=("$env_file")
+  fi
+done
+
+env_yaml="$(mktemp --suffix=.yml)"
+conda-merge environment.yml "${additional_env_files[@]}" | tee "$env_yaml"
+mamba env update --name ibis --file "$env_yaml"
diff --git a/ci/omniscidb.conf b/ci/omniscidb.conf
diff --git a/ci/recipe/meta.yaml b/ci/recipe/meta.yaml
@@ -16,10 +16,9 @@ source:
   url: https://github.com/ibis-project/ibis/archive/{{ version }}.tar.gz
 
 build:
-  number: 1
+  number: 0
   script: {{ PYTHON }} -m pip install . --no-deps --ignore-installed --no-cache-dir -vvv
-  # uncomment noarch when pymapd and pyspark issues are fixed for py38
-  # noarch: python
+  skip: true  # [py<37]
 
 requirements:
   host:
@@ -28,24 +27,23 @@ requirements:
     - setuptools
 
   run:
+    - cached_property
     - clickhouse-driver >=0.1.3
     - clickhouse-cityhash  # [not win]
     - clickhouse-sqlalchemy
     - geoalchemy2
     - geopandas
-    - google-cloud-bigquery-core >=1.12.0,<1.24.0dev
     - graphviz
     - impyla >=0.15.0
     - lz4
     - multipledispatch >=0.6
     - numpy >=1.15
     - pandas >=0.25.3
+    - parsy
     - psycopg2
     - pyarrow >=0.15
-    - pydata-google-auth
-    - pymapd 0.24  # [py<38]
     - pymysql
-    - pyspark >=2.4.3  # [py<38]
+    - pyspark >=2.4.3
     - pytables >=3.0.0
     - python
     - python-graphviz
@@ -59,22 +57,20 @@ requirements:
     - thrift >=0.11
     - thriftpy2
     - toolz
+    - tzlocal <3 # not directly needed, but need to pin since 3.0 is broken
 
 test:
   imports:
     - ibis
-    - ibis.backends.bigquery
     - ibis.backends.clickhouse
     - ibis.backends.csv
     - ibis.backends.parquet
     - ibis.backends.hdf5
     - ibis.backends.impala
     - ibis.backends.mysql
-    - ibis.backends.omniscidb  # [py<38]
     - ibis.backends.pandas
     - ibis.backends.postgres
-    - ibis.backends.pyspark  # [py<38]
-    - ibis.backends.spark
+    - ibis.backends.pyspark
     - ibis.backends.sqlite
 
 about:
@@ -86,9 +82,6 @@ about:
 
 extra:
   recipe-maintainers:
-    - cpcloud
-    - mariusvniekerk
-    - wesm
-    - kszucs
-    - xmnlab
     - jreback
+    - xmnlab
+    - datapythonista
diff --git a/ci/run_tests.sh b/ci/run_tests.sh
@@ -1,22 +1,26 @@
-#!/bin/bash -e
+#!/usr/bin/env bash
 # Run the Ibis tests. Two environment variables are considered:
 # - PYTEST_BACKENDS: Space-separated list of backends to run
-# - PYTEST_EXPRESSION: Marker expression, for example "not udf"
 
-TESTS_DIRS="ibis/tests"
-for BACKEND in $PYTEST_BACKENDS; do
-    if [[ -d ibis/$BACKEND/tests ]]; then
-        TESTS_DIRS="$TESTS_DIRS ibis/$BACKEND/tests"
+set -eo pipefail
+
+TESTS_DIRS=()
+
+if [ -n "$PYTEST_BACKENDS" ]; then
+    TESTS_DIRS+=("ibis/backends/tests")
+fi
+
+for backend in $PYTEST_BACKENDS; do
+    backend_test_dir="ibis/backends/$backend/tests"
+    if [ -d "$backend_test_dir" ]; then
+        TESTS_DIRS+=("$backend_test_dir")
     fi
 done
 
-echo "TESTS_DIRS: $TESTS_DIRS"
-echo "PYTEST_EXPRESSION: $PYTEST_EXPRESSION"
-
+set -x
 
-pytest $TESTS_DIRS \
-    -m "${PYTEST_EXPRESSION}" \
+pytest "${TESTS_DIRS[@]}" \
     -ra \
     --junitxml=junit.xml \
     --cov=ibis \
-    --cov-report=xml:coverage.xml
+    --cov-report=xml:coverage.xml "$@"
diff --git a/ci/schema/bigquery.sql b/ci/schema/bigquery.sql
diff --git a/ci/schema/omniscidb.sql b/ci/schema/omniscidb.sql
diff --git a/ci/setup_env.sh b/ci/setup_env.sh
diff --git a/conftest.py b/conftest.py
diff --git a/dev/merge-pr.py b/dev/merge-pr.py
@@ -64,11 +64,11 @@ def merge_pr(
         "{GITHUB_API_BASE}/pulls/{pr_num:d}/merge".format(
             GITHUB_API_BASE=GITHUB_API_BASE, pr_num=pr_num
         ),
-        json=dict(
-            commit_title=commit_title,
-            commit_message=commit_message,
-            merge_method=merge_method,
-        ),
+        json={
+            'commit_title': commit_title,
+            'commit_message': commit_message,
+            'merge_method': merge_method,
+        },
         auth=(github_user, password),
     )
     status_code = resp.status_code

diff --git a/docs/source/api.rst b/docs/source/api.rst
@@ -39,6 +39,25 @@ These methods are available directly in the ``ibis`` module namespace.
    trailing_range_window
    random
 
+.. _api.basebackend:
+
+Backend methods
+------------------
+
+.. currentmodule:: ibis.backends.base
+
+.. autosummary::
+   :toctree: generated/
+
+   BaseBackend.connect
+   BaseBackend.database
+   BaseBackend.current_database
+   BaseBackend.list_tables
+   BaseBackend.table
+   BaseBackend.version
+   BaseBackend.compile
+   BaseBackend.execute
+
 .. _api.expr:
 
 General expression methods
@@ -66,6 +85,7 @@ Table methods
    :toctree: generated/
 
    TableExpr.aggregate
+   TableExpr.asof_join
    TableExpr.count
    TableExpr.distinct
    TableExpr.drop
@@ -229,6 +249,9 @@ Scalar or column methods
 
    IntegerValue.convert_base
    IntegerValue.to_timestamp
+   IntegerColumn.bit_and
+   IntegerColumn.bit_or
+   IntegerColumn.bit_xor
 
 .. _api.string:
 
@@ -419,6 +442,22 @@ Decimal methods
    DecimalValue.precision
    DecimalValue.scale
 
+.. _api.struct:
+
+Struct methods
+-----------------
+
+Scalar or column methods
+~~~~~~~~~~~~~~~~~~~~~~~~
+
+Values in a ``StructValue`` can be accessed using indexing, e.g. ``struct_expr['my_col']``. See :meth:`StructValue.__getitem__`.
+
+.. autosummary::
+   :toctree: generated/
+
+   StructValue.destructure
+   StructValue.__getitem__
+
 .. _api.geospatial:
 
 Geospatial methods
@@ -450,7 +489,7 @@ Scalar or column methods
    GeoSpatialValue.distance
    GeoSpatialValue.end_point
    GeoSpatialValue.envelope
-   GeoSpatialValue.equals
+   GeoSpatialValue.geo_equals
    GeoSpatialValue.geometry_n
    GeoSpatialValue.geometry_type
    GeoSpatialValue.intersection

diff --git a/docs/source/backends/bigquery.rst b/docs/source/backends/bigquery.rst
diff --git a/docs/source/backends/clickhouse.rst b/docs/source/backends/clickhouse.rst
@@ -7,7 +7,7 @@ Install dependencies for Ibis's Clickhouse dialect(minimal supported version is
 
 ::
 
-  pip install ibis-framework[clickhouse]
+  pip install 'ibis-framework[clickhouse]'
 
 Create a client by passing in database connection parameters such as ``host``,
 ``port``, ``database``, and ``user`` to :func:`ibis.clickhouse.connect`:
@@ -30,11 +30,10 @@ Use ``ibis.clickhouse.connect`` to create a client.
 .. autosummary::
    :toctree: ../generated/
 
-   connect
-   ClickhouseClient.close
-   ClickhouseClient.exists_table
-   ClickhouseClient.exists_database
-   ClickhouseClient.get_schema
-   ClickhouseClient.set_database
-   ClickhouseClient.list_databases
-   ClickhouseClient.list_tables
+   Backend.connect
+   Backend.close
+   Backend.exists_table
+   Backend.exists_database
+   Backend.get_schema
+   Backend.list_databases
+   Backend.list_tables
diff --git a/docs/source/backends/dask.rst b/docs/source/backends/dask.rst
@@ -0,0 +1,4 @@
+`Dask <https://dask.org/>`_
+===========================
+
+Dask backend is currently experimental.
diff --git a/docs/source/backends/impala.rst b/docs/source/backends/impala.rst
diff --git a/docs/source/backends/index.rst b/docs/source/backends/index.rst
@@ -15,11 +15,10 @@ For more information on a specific backend, check the next backend pages:
    postgres
    mysql
    impala
-   omnisci
-   bigquery
    clickhouse
-   spark
+   pyspark
    pandas
+   dask
 
 
 .. _classes_of_backends:
@@ -44,7 +43,6 @@ string to the database through a driver API.
 - `Google BigQuery <https://cloud.google.com/bigquery/>`_
 - `Hadoop Distributed File System (HDFS) <https://hadoop.apache.org/>`_
 - `OmniSciDB <https://www.omnisci.com/>`_
-- `PySpark/Spark SQL <https://spark.apache.org/sql/>`_ (Experimental)
 
 .. _expression_generating_backends:
 
@@ -61,7 +59,7 @@ dependencies).
 
 - `PostgreSQL <https://www.postgresql.org/>`_
 - `SQLite <https://www.sqlite.org/>`_
-- `MySQL <https://www.mysql.com/>`_ (Experimental)
+- `MySQL <https://www.mysql.com/>`_
 
 .. _direct_execution_backends:
 
@@ -72,3 +70,5 @@ backend. A full description of the implementation can be found in the module
 docstring of the pandas backend located in ``ibis/backends/pandas/core.py``.
 
 - `Pandas <http://pandas.pydata.org/>`_
+- `PySpark <https://spark.apache.org/sql/>`_
+- `Dask <https://dask.org/>`_ (Experimental)
diff --git a/docs/source/backends/mysql.rst b/docs/source/backends/mysql.rst
@@ -7,7 +7,7 @@ Install dependencies for Ibis's MySQL dialect:
 
 ::
 
-  pip install ibis-framework[mysql]
+  pip install 'ibis-framework[mysql]'
 
 Create a client by passing a connection string or individual parameters to
 :func:`ibis.mysql.connect`:
@@ -36,8 +36,8 @@ create a client.
 .. autosummary::
    :toctree: ../generated/
 
-   connect
-   MySQLClient.database
-   MySQLClient.list_databases
-   MySQLClient.list_tables
-   MySQLClient.table
+   Backend.connect
+   Backend.database
+   Backend.list_databases
+   Backend.list_tables
+   Backend.table
diff --git a/docs/source/backends/omnisci.rst b/docs/source/backends/omnisci.rst
diff --git a/docs/source/backends/postgres.rst b/docs/source/backends/postgres.rst
@@ -7,7 +7,7 @@ Install dependencies for Ibis's PostgreSQL dialect:
 
 ::
 
-  pip install ibis-framework[postgres]
+  pip install 'ibis-framework[postgres]'
 
 Create a client by passing a connection string to the ``url`` parameter or
 individual parameters to :func:`ibis.postgres.connect`:
@@ -39,8 +39,8 @@ create a client.
 .. autosummary::
    :toctree: ../generated/
 
-   connect
-   PostgreSQLClient.database
-   PostgreSQLClient.list_tables
-   PostgreSQLClient.list_databases
-   PostgreSQLClient.table
+   Backend.connect
+   Backend.database
+   Backend.list_tables
+   Backend.list_databases
+   Backend.table
diff --git a/docs/source/backends/pyspark.rst b/docs/source/backends/pyspark.rst
@@ -0,0 +1,36 @@
+.. _install.pyspark:
+
+`PySpark <https://spark.apache.org/sql/>`_
+====================================================
+
+Install dependencies for Ibis's PySpark dialect:
+
+::
+
+  pip install 'ibis-framework[pyspark]'
+
+.. note::
+
+   When using the PySpark backend with PySpark 2.4.x and pyarrow >= 0.15.0, you
+   need to set ``ARROW_PRE_0_15_IPC_FORMAT=1``. See `here
+   <https://spark.apache.org/docs/latest/api/python/user_guide/arrow_pandas.html#compatibility-setting-for-pyarrow-0-15-0-and-spark-2-3-x-2-4-x>`_
+   for details
+
+.. _api.pyspark:
+
+PySpark client
+~~~~~~~~~~~~~~
+.. currentmodule:: ibis.backends.pyspark
+
+The PySpark client is accessible through the ``ibis.pyspark`` namespace.
+
+Use ``ibis.pyspark.connect`` to create a client.
+
+.. autosummary::
+   :toctree: ../generated/
+
+   Backend.connect
+   Backend.database
+   Backend.list_databases
+   Backend.list_tables
+   Backend.table
diff --git a/docs/source/backends/spark.rst b/docs/source/backends/spark.rst
diff --git a/docs/source/backends/sqlite.rst b/docs/source/backends/sqlite.rst
@@ -7,7 +7,7 @@ Install dependencies for Ibis's SQLite dialect:
 
 ::
 
-  pip install ibis-framework[sqlite]
+  pip install 'ibis-framework[sqlite]'
 
 Create a client by passing a path to a SQLite database to
 :func:`ibis.sqlite.connect`:
@@ -33,8 +33,8 @@ Use ``ibis.sqlite.connect`` to create a SQLite client.
 .. autosummary::
    :toctree: ../generated/
 
-   connect
-   SQLiteClient.attach
-   SQLiteClient.database
-   SQLiteClient.list_tables
-   SQLiteClient.table
+   Backend.connect
+   Backend.attach
+   Backend.database
+   Backend.list_tables
+   Backend.table
diff --git a/docs/source/conf.py b/docs/source/conf.py
@@ -1,4 +1,3 @@
-# -*- coding: utf-8 -*-
 #
 # Ibis documentation build configuration file, created by
 # sphinx-quickstart on Wed Jun 10 11:06:29 2015.
@@ -68,7 +67,7 @@
 
 # General information about the project.
 project = 'Ibis'
-copyright = '{}, Ibis Developers'.format(datetime.date.today().year)
+copyright = f'{datetime.date.today().year}, Ibis Developers'
 
 # The version info for the project you're documenting, acts as replacement for
 # |version| and |release|, also used in various other places throughout the

diff --git a/docs/source/index.rst b/docs/source/index.rst
@@ -64,7 +64,7 @@ Ibis can also be installed with Kerberos support for its HDFS functionality:
 
 ::
 
-  pip install ibis-framework[kerberos]
+  pip install 'ibis-framework[kerberos]'
 
 Some platforms will require that you have Kerberos installed to build properly.
 

diff --git a/docs/source/release/index.rst b/docs/source/release/index.rst
@@ -12,6 +12,77 @@ Release Notes
    These release notes are for versions of ibis **1.0 and later**. Release
    notes for pre-1.0 versions of ibis can be found at :doc:`release-pre-1.0`
 
+* :support:`2678` Improvement of the backend API. The former `Client` subclasses have been replaced by a `Backend` class that must
+  subclass `ibis.backends.base.BaseBackend`. The `BaseBackend` class contains abstract methods for the minimum subset of methods that
+  backends must implement, and their signatures have been standardized across backends. The Ibis compiler has been refactored, and
+  backends don't need to implement all compiler classes anymore if the default works for them. Only a subclass of
+  `ibis.backends.base.sql.compiler.Compiler` is now required. Backends now need to register themselves as entry points.
+* :support:`2905` Deprecate `exists_table(table)` in favor of `table in list_tables()`
+* :bug:`2991` Fix data races in impala connection pool accounting
+* :bug:`2985` Fix null literal compilation in the Clickhouse backend
+* :bug:`2984` Fix order of limit and offset parameters in the Clickhouse backend
+* :support:`2977` Remove handwritten type parser; parsing errors that were previously `IbisTypeError` are now `parsy.ParseError`. `parsy` is now a hard requirement.
+* :support:`2962` Methods `current_database` and `list_databases` raise an exception for backends that do not support databases
+* :bug:`2956` Replace `equals` operation for geospatial datatype to `geo_equals` 
+* :support:`2913` Method `set_database` has been deprecated, in favor of creating a new connection to a different database
+* :feature:`2938` Serialization-deserialization of Node via pickle is now byte compatible between different processes
+* :support:`2914` Removed `log` method of clients, in favor of `verbose_log` option
+* :feature:`2916` Support joining on different columns in ClickHouse backend
+* :feature:`2908` Support summarization of empty data in Pandas backend
+* :support:`2883` Output of `Client.version` returned as a string, instead of a setuptools `Version`
+* :feature:`2882` Unify implementation of fillna and isna in Pyspark backend
+* :support:`2862` Deprecated `list_schemas` in SQLAlchemy backends in favor of `list_databases`
+* :bug:`2829` Fix .drop(fields). The argument can now be either a list of strings or a string.
+* :feature:`2873` Support binary operation with Timedelta in Pyspark backend
+* :support:`2865` Deprecated `ibis.<backend>.verify()` in favor of capturing exception in `ibis.<backend>.compile()`
+* :bug:`2845` Fix projection on differences and intersections for SQL backends
+* :feature:`2839`: Add `group_concat` operation for Clickhouse backend
+* :bug:`2827` Backends are loaded in a lazy way, so third-party backends can import Ibis without circular imports
+* :bug:`2830` Disable aggregation optimization due to N squared performance
+* :bug:`2821` Fix `.cast()` to array outputting list instead of np.array in Pandas backend
+* :bug:`2820` Fix aggregation with mixed reduction datatypes (array + scalar) on Dask backend
+* :feature:`2808` Support comparison of ColumnExpr to timestamp literal
+* :support:`2789` Simplification of data fetching. Backends don't need to implement `Query` anymore
+* :feature:`2805` Make op schema a cached property
+* :feature:`2613` :feature:`2778` Implement `.insert()` for SQLAlchemy backends
+* :feature:`2792` Infer categorical and decimal Series to more specific Ibis types in Pandas backend
+* :feature:`2790` Add `startswith` and `endswith` operations
+* :feature:`2776` :feature:`2797` Allow more flexible return type for UDFs
+* :feature:`2779` Implement Clip in the Pyspark backend
+* :bug:`2770` Fix error when using reduction UDF that returns np.array in a grouped aggregation
+* :feature:`2753` Use `ndarray` as array representation in Pandas backend
+* :support:`2665` Move BigQuery backend to a `separate repository <https://github.com/ibis-project/ibis-bigquery>`_.
+  The backend will be released separately, use `pip install ibis-bigquery` or `conda install ibis-bigquery` to
+  install it, and then use as before.
+* :bug:`2712` Fix time context trimming error for multi column udfs in pandas backend
+* :bug:`2710` Fix error during compilation of range_window in base_sql backends (:issue:`2608`)
+* :feature:`2687` Support Spark filter with window operation
+* :bug:`2696` Fix wrong row indexing in the result for 'window after filter' for timecontext adjustment
+* :bug:`2702` Fix `aggregate` exploding the output of Reduction ops that return a list/ndarray
+* :bug:`2693` Fix issues with context adjustment for filter with PySpark backend
+* :support:`2689` Supporting SQLAlchemy 1.4, and requiring minimum 1.3
+* :support:`2680` Namespace time_col config, fix type check for trim_with_timecontext for pandas window execution
+* :feature:`2646` Support context adjustment for udfs for pandas backend
+* :feature:`2655` Add `auth_local_webserver`, `auth_external_data`, and
+  `auth_cache` parameters to BigQuery connect method. Set
+  `auth_local_webserver` to use a local server instead of copy-pasting an
+  authorization code. Set `auth_external_data` to true to request additional
+  scopes required to query Google Drive and Sheets. Set `auth_cache` to
+  `reauth` or `none` to force reauthentication.
+* :bug:`2657` Add temporary struct col in pyspark backend to ensure that UDFs are executed only once
+* :bug:`2588` Fix BigQuery connect bug that ignored project ID parameter
+* :bug:`2636` Fix overwrite logic to account for DestructColumn inside mutate API
+* :feature:`2641` Add `bit_and`, `bit_or`, and `bit_xor` integer column aggregates (BigQuery and MySQL backends)
+* :feature:`2379` Backends are defined as entry points
+* :bug:`2635` Fix fusion optimization bug that incorrectly changes operation order
+* :feature:`2615` Add `ibis.array` for creating array expressions
+* :feature:`2607` Implement Not operation in PySpark backend
+* :feature:`2610` Added support for case/when in PySpark backend
+* :bug:`2610` Fixes a NPE issue with substr in PySpark backend
+* :feature:`2603` Add support for np.array as literals for backends that already support lists as literals
+* :bug:`2354` Fixes binary data type translation into BigQuery bytes data type
+* :bug:`2577` Make StructValue picklable
+* :support:`2505` Remove deprecated `ibis.HDFS`, `ibis.WebHDFS` and `ibis.hdfs_connect`
 * :feature:`2514` Add Struct.from_dict
 * :feature:`2310` Add hash and hashbytes support for BigQuery backend
 * :feature:`2511` Support reduction UDF without groupby to return multiple columns for Pandas backend

diff --git a/docs/source/release/release-pre-1.0.rst b/docs/source/release/release-pre-1.0.rst
@@ -20,8 +20,8 @@ New Features
 * Splat args into Node subclasses instead of requiring a list (:ghissue:`969`)
 * Add support for ``UNION`` in the BigQuery backend (:ghissue:`1408`,
   :ghissue:`1409`)
-* Support for writing UDFs in BigQuery (:ghissue:`1377`). See :ref:`the BigQuery
-  UDF docs <udf.bigquery>` for more details.
+* Support for writing UDFs in BigQuery (:ghissue:`1377`). See the BigQuery
+  UDF docs for more details.
 * Support for cross-project expressions in the BigQuery backend.
   (:ghissue:`1427`, :ghissue:`1428`)
 * Add ``strftime`` and ``to_timestamp`` support for BigQuery (:ghissue:`1422`,

diff --git a/docs/source/tutorial/01-Introduction-to-Ibis.ipynb b/docs/source/tutorial/01-Introduction-to-Ibis.ipynb
@@ -85,7 +85,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Note that if you installed Ibis with `pip` instead of `conda`, you may need to install the SQLite backend separately with `pip install ibis-framework[sqlite]`.\n",
+    "Note that if you installed Ibis with `pip` instead of `conda`, you may need to install the SQLite backend separately with `pip install 'ibis-framework[sqlite]'`.\n",
     "\n",
     "### Exploring the data\n",
     "\n",

diff --git a/docs/source/tutorial/04-More-Value-Expressions.ipynb b/docs/source/tutorial/04-More-Value-Expressions.ipynb
@@ -20,13 +20,12 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "import ibis\n",
     "import os\n",
-    "hdfs_port = os.environ.get('IBIS_WEBHDFS_PORT', 50070)\n",
-    "hdfs = ibis.hdfs_connect(host='impala', port=hdfs_port)\n",
-    "con = ibis.impala.connect(host='impala', database='ibis_testing',\n",
-    "                          hdfs_client=hdfs)\n",
-    "ibis.options.interactive = True"
+    "import ibis\n",
+    "\n",
+    "ibis.options.interactive = True\n",
+    "\n",
+    "connection = ibis.sqlite.connect(os.path.join('data', 'geography.db'))"
    ]
   },
   {
@@ -57,8 +56,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "table = con.table('functional_alltypes')\n",
-    "table.string_col.cast('double').sum()"
+    "countries = connection.table('countries')\n",
+    "countries\n",
+    "connection.table('gdp')"
    ]
   },
   {
@@ -67,7 +67,17 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "table.string_col.cast('decimal(12,2)').sum()"
+    "countries = connection.table('countries')\n",
+    "countries.population.cast('float').sum()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "countries.area_km2.cast('int32').sum()"
    ]
   },
   {
@@ -86,15 +96,18 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "expr = (table.string_col\n",
+    "expr = (countries.continent\n",
     "        .case()\n",
-    "        .when('4', 'fee')\n",
-    "        .when('7', 'fi')\n",
-    "        .when('1', 'fo')\n",
-    "        .when('0', 'fum')\n",
-    "        .else_(table.string_col)\n",
+    "        .when('AF', 'Africa')\n",
+    "        .when('AN', 'Antarctica')\n",
+    "        .when('AS', 'Asia')\n",
+    "        .when('EU', 'Europe')\n",
+    "        .when('NA', 'North America')\n",
+    "        .when('OC', 'Oceania')\n",
+    "        .when('SA', 'South America')\n",
+    "        .else_(countries.continent)\n",
     "        .end()\n",
-    "        .name('new_strings'))\n",
+    "        .name('continent_name'))\n",
     "\n",
     "expr.value_counts()"
    ]
@@ -112,12 +125,16 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "expr = (table.string_col\n",
+    "expr = (countries.continent\n",
     "        .case()\n",
-    "        .when('4', 'fee')\n",
-    "        .when('7', 'fi')\n",
+    "        .when('AF', 'Africa')\n",
+    "        .when('AS', 'Asia')\n",
+    "        .when('EU', 'Europe')\n",
+    "        .when('NA', 'North America')\n",
+    "        .when('OC', 'Oceania')\n",
+    "        .when('SA', 'South America')\n",
     "        .end()\n",
-    "        .name('with_nulls'))\n",
+    "        .name('continent_name_with_nulls'))\n",
     "\n",
     "expr.value_counts()"
    ]
@@ -136,12 +153,13 @@
    "outputs": [],
    "source": [
     "expr = (ibis.case()\n",
-    "        .when(table.int_col > 5, table.bigint_col * 2)\n",
-    "        .when(table.int_col > 2, table.bigint_col)\n",
-    "        .else_(table.int_col)\n",
-    "        .end())\n",
+    "        .when(countries.population > 25_000_000, 'big')\n",
+    "        .when(countries.population < 5_000_000, 'small')\n",
+    "        .else_('medium')\n",
+    "        .end()\n",
+    "        .name('size'))\n",
     "\n",
-    "table['id', 'int_col', 'bigint_col', expr.name('case_result')].limit(20)"
+    "countries['name', 'population', expr].limit(10)"
    ]
   },
   {
@@ -157,11 +175,11 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "expr = ((table.int_col > 5)\n",
-    "        .ifelse(table.bigint_col / 2, table.bigint_col * 2)\n",
-    "        .name('ifelse_result'))\n",
+    "expr = ((countries.continent == 'AS')\n",
+    "        .ifelse('Asia', 'Not Asia')\n",
+    "        .name('is_asia'))\n",
     "\n",
-    "table['int_col', 'bigint_col', expr].limit(10)"
+    "countries['name', 'continent', expr].limit(10)"
    ]
   },
   {
@@ -183,8 +201,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "bool_clause = table.string_col.notin(['1', '4', '7'])\n",
-    "table[bool_clause].string_col.value_counts()"
+    "is_america = countries.continent.isin(['NA', 'SA'])\n",
+    "countries[is_america].continent.value_counts()"
    ]
   },
   {
@@ -200,9 +218,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "top_strings = table.string_col.value_counts().limit(3).string_col\n",
-    "top_filter = table.string_col.isin(top_strings)\n",
-    "expr = table[top_filter]\n",
+    "top_continents = countries.continent.value_counts().limit(3).continent\n",
+    "top_continents_filter = countries.continent.isin(top_continents)\n",
+    "expr = countries[top_continents_filter]\n",
     "\n",
     "expr.count()"
    ]
@@ -220,7 +238,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "table[table.string_col.topk(3)].count()"
+    "countries.continent.topk(3)"
    ]
   },
   {
@@ -245,13 +263,13 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "expr = (table.string_col\n",
+    "expr = (countries.continent\n",
     "        .case()\n",
-    "        .when('4', 'fee')\n",
-    "        .when('7', 'fi')\n",
-    "        .when('1', 'fo')\n",
+    "        .when('AF', 'Africa')\n",
+    "        .when('EU', 'Europe')\n",
+    "        .when('AS', 'Asia')\n",
     "        .end()\n",
-    "        .name('new_strings'))\n",
+    "        .name('top_continent_name'))\n",
     "\n",
     "expr.isnull().value_counts()"
    ]
@@ -269,7 +287,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "expr2 = expr.isnull().ifelse('was null', expr).name('strings')\n",
+    "expr2 = expr.isnull().ifelse('Other continent', expr).name('continent')\n",
     "expr2.value_counts()"
    ]
   },
@@ -289,7 +307,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "table['int_col', 'bigint_col'].distinct()"
+    "countries['continent'].distinct()"
    ]
   },
   {
@@ -298,7 +316,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "table.string_col.distinct()"
+    "countries.continent.distinct()"
    ]
   },
   {
@@ -314,9 +332,10 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "metric = (table.bigint_col\n",
+    "metric = (countries.continent\n",
     "          .distinct().count()\n",
-    "          .name('unique_bigints'))"
+    "          .name('num_continents'))\n",
+    "metric"
    ]
   },
   {
@@ -332,7 +351,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "table.string_col.nunique()"
+    "countries.continent.nunique()"
    ]
   },
   {
@@ -342,7 +361,7 @@
     "## String operations\n",
     "\n",
     "\n",
-    "What's supported is pretty basic right now. We intend to support the full gamut of regular expression munging with a nice API, though in some cases some work will be required on Impala's backend to support everything. "
+    "What's supported is pretty basic right now. We intend to support the full gamut of regular expression munging with a nice API, though in some cases some work will be required on SQLite's backend to support everything. "
    ]
   },
   {
@@ -351,8 +370,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "nation = con.table('tpch_nation')\n",
-    "nation.limit(5)"
+    "countries[['name']].limit(5)"
    ]
   },
   {
@@ -368,8 +386,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "expr = nation.n_name.lower().left(1).name('first_letter')\n",
-    "expr.value_counts().sort_by(('count', False))"
+    "expr = countries.name.lower().left(1).name('first_letter')\n",
+    "expr.value_counts().sort_by(('count', False)).limit(10)"
    ]
   },
   {
@@ -389,7 +407,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "nation[nation.n_name.like('%GE%')]"
+    "countries[countries.name.like('%GE%')].name"
    ]
   },
   {
@@ -398,7 +416,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "nation[nation.n_name.lower().rlike('.*ge.*')]"
+    "countries[countries.name.lower().rlike('.*ge.*')].name"
    ]
   },
   {
@@ -407,7 +425,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "nation[nation.n_name.lower().contains('ge')]"
+    "countries[countries.name.lower().contains('ge')].name"
    ]
   },
   {
@@ -430,9 +448,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "table = con.table('functional_alltypes')\n",
+    "independence = connection.table('independence')\n",
     "\n",
-    "table[table.timestamp_col, table.timestamp_col.minute().name('minute')].limit(10)"
+    "independence[independence.independence_date, independence.independence_date.month().name('month')].limit(10)"
    ]
   },
   {
@@ -449,11 +467,11 @@
    "outputs": [],
    "source": [
     "def get_field(f):\n",
-    "    return getattr(table.timestamp_col, f)().name(f)\n",
+    "    return getattr(independence.independence_date, f)().name(f)\n",
     "\n",
-    "fields = ['year', 'month', 'day', 'hour', 'minute', 'second', 'millisecond']\n",
-    "projection = [table.timestamp_col] + [get_field(x) for x in fields]\n",
-    "table[projection].limit(10)"
+    "fields = ['year', 'month', 'day']  # datetime fields can also use: 'hour', 'minute', 'second', 'millisecond'\n",
+    "projection = [independence.independence_date] + [get_field(x) for x in fields]\n",
+    "independence[projection].limit(10)"
    ]
   },
   {
@@ -469,7 +487,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "table[table.timestamp_col.min(), table.timestamp_col.max(), table.count().name('nrows')]"
+    "independence[independence.independence_date.min(),\n",
+    "             independence.independence_date.max(),\n",
+    "             independence.count().name('nrows')].distinct()"
    ]
   },
   {
@@ -478,27 +498,14 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "table[table.timestamp_col < '2010-01-01'].count()"
+    "independence[independence.independence_date > '2000-01-01'].count()"
    ]
   },
   {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "table[table.timestamp_col < \n",
-    "      (ibis.timestamp('2010-01-01') + ibis.interval(months=3))].count()"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
+   "cell_type": "markdown",
    "metadata": {},
-   "outputs": [],
    "source": [
-    "expr = (table.timestamp_col + ibis.interval(days=1) + ibis.interval(hours=4)).name('offset')\n",
-    "table[table.timestamp_col, expr, ibis.now().name('current_time')].limit(10)"
+    "Some backends support adding offsets, for example `independence.independence_date + ibis.interval(days=1)` or `ibis.now() - independence.independence_date`."
    ]
   }
  ],
@@ -518,9 +525,9 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.6.7"
+   "version": "3.9.1"
   }
  },
  "nbformat": 4,
- "nbformat_minor": 1
+ "nbformat_minor": 4
 }
diff --git a/docs/source/tutorial/05-IO-Create-Insert-External-Data.ipynb b/docs/source/tutorial/05-IO-Create-Insert-External-Data.ipynb
diff --git a/docs/source/tutorial/06-Advanced-Topics-ComplexFiltering.ipynb b/docs/source/tutorial/06-Advanced-Topics-ComplexFiltering.ipynb
@@ -0,0 +1,208 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# Advanced Topics: Additional Filtering\n",
+    "\n",
+    "The filtering examples we've shown to this point have been pretty simple, either comparisons between columns or fixed values, or set filter functions like `isin` and `notin`. \n",
+    "\n",
+    "Ibis supports a number of richer analytical filters that can involve one or more of:\n",
+    "\n",
+    "- Aggregates computed from the same or other tables\n",
+    "- Conditional aggregates (in SQL-speak these are similar to \"correlated subqueries\")\n",
+    "- \"Existence\" set filters (equivalent to the SQL `EXISTS` and `NOT EXISTS` keywords)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Setup"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import ibis\n",
+    "\n",
+    "ibis.options.interactive = True\n",
+    "\n",
+    "connection = ibis.sqlite.connect(os.path.join('data', 'geography.db'))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Using scalar aggregates in filters"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "countries = connection.table('countries')\n",
+    "countries.limit(5)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "We could always compute some aggregate value from the table and use that in another expression, or we can use a data-derived aggregate in the filter. Take the average of a column. For example the average of countries size:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "countries.area_km2.mean()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "You can use this expression as a substitute for a scalar value in a filter, and the execution engine will combine everything into a single query rather than having to access the database multiple times. For example, we want to filter European countries larger than the average country size in the world. See how most countries in Europe are smaller than the world average:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "cond = countries.area_km2 > countries.area_km2.mean()\n",
+    "expr = countries[(countries.continent == 'EU') & cond]\n",
+    "expr"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Conditional aggregates\n",
+    "\n",
+    "\n",
+    "Suppose that we wish to filter using an aggregate computed conditional on some other expressions holding true.\n",
+    "\n",
+    "For example, we want to filter European countries larger than the average country size, but this time of the average in Africa. African countries have an smaller size compared to the world average, and France gets into the list:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "conditional_avg = countries[countries.continent == 'AF'].area_km2.mean()\n",
+    "countries[(countries.continent == 'EU') & (countries.area_km2 > conditional_avg)]"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## \"Existence\" filters\n",
+    "\n",
+    "\n",
+    "Some filtering involves checking for the existence of a particular value in a column of another table, or amount the results of some value expression. This is common in many-to-many relationships, and can be performed in numerous different ways, but it's nice to be able to express it with a single concise statement and let Ibis compute it optimally.\n",
+    "\n",
+    "An example could be finding all countries that had **any** year with a higher GDP than 3 trillion US dollars:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "gdp = connection.table('gdp')\n",
+    "gdp"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "cond = ((gdp.country_code == countries.iso_alpha3) &\n",
+    "        (gdp.value > 3e12)).any()\n",
+    "\n",
+    "countries[cond]['name']"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Note how this is different than a join between `countries` and `gdp`, which would return one row per year. The method `.any()` is equivalent to filtering with a subquery."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Filtering in aggregations\n",
+    "\n",
+    "\n",
+    "Suppose that you want to compute an aggregation with a subset of the data for _only one_ of the metrics / aggregates in question, and the complete data set with the other aggregates. Most aggregation functions are thus equipped with a `where` argument. Let me show it to you in action:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "arctic = countries.name.isin(['United States',\n",
+    "                              'Canada',\n",
+    "                              'Finland',\n",
+    "                              'Greenland',\n",
+    "                              'Iceland',\n",
+    "                              'Norway',\n",
+    "                              'Russia',\n",
+    "                              'Sweden'])\n",
+    "\n",
+    "metrics = [countries.count().name('# countries'),\n",
+    "           countries.population.sum().name('total population'),\n",
+    "           countries.population.sum(where=arctic).name('population arctic countries')]\n",
+    "\n",
+    "(countries.groupby(countries.continent)\n",
+    "          .aggregate(metrics))"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.1"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}
diff --git a/docs/source/tutorial/06-Advanced-Topics-TopK-SelfJoins.ipynb b/docs/source/tutorial/06-Advanced-Topics-TopK-SelfJoins.ipynb
-Original file line number
+Diff line change
@@ Expand Up @@
     ::
-      pip install ibis-framework[kerberos]
+      pip install 'ibis-framework[kerberos]'
     Some platforms will require that you have Kerberos installed to build properly.
@@ Expand Down @@