dev: simplify

.dockerignore

@@ -1,2 +1,3 @@
 .venv
 ml-25m*
+dev/ml-25m*

Dockerfile

@@ -0,0 +1,11 @@
+FROM python:3.10
+RUN apt-get update && apt-get install -y \
+    python3-dev libpq-dev wget unzip \
+    python3-setuptools gcc bc
+RUN pip install --no-cache-dir poetry==1.1.13
+COPY . /app
+WORKDIR /app
+# For now while we are in heavy development we install the latest with Poetry 
+# and execute directly with Poetry. Later, we'll move to the released Pip package.
+RUN poetry install -E preql -E mysql -E pgsql -E snowflake
+ENTRYPOINT ["poetry", "run", "python3", "-m", "data_diff"]

README.md

@@ -1,27 +1,74 @@
 # Data Diff
 
-A cross-database, efficient diff between mostly-similar database tables.
+A cross-database, efficient diff using checksums between mostly-similar database
+tables.
 
-Use cases:
+- Validate that a table was copied properly
+- Be alerted before your customer finds out, or your report is wrong
+- Validate that your replication mechnism is working correctly
+- Find changes between two versions of the same table
 
-- Quickly validate that a table was copied correctly
+It uses a bisection algorithm to efficiently check if e.g. a table is the same
+between MySQL and Postgres, or Postgres and Snowflake, or MySQL and RDS!
 
-- Find changes between two versions of the same table
+```python
+$ data-diff postgres:/// Original  postgres:/// Original_1diff  -v --bisection-factor=4
+[16:55:19] INFO - Diffing tables of size 25000095 and 25000095 | segments: 4, bisection threshold: 1048576.
+[16:55:36] INFO - Diffing segment 0/4 of size 8333364 and 8333364
+[16:55:45] INFO - . Diffing segment 0/4 of size 2777787 and 2777787
+[16:55:52] INFO - . . Diffing segment 0/4 of size 925928 and 925928
+[16:55:54] INFO - . . . Diff found 2 different rows.
++ (20000, 942013020)
+- (20000, 942013021)
+[16:55:54] INFO - . . Diffing segment 1/4 of size 925929 and 925929
+[16:55:55] INFO - . . Diffing segment 2/4 of size 925929 and 925929
+[16:55:55] INFO - . . Diffing segment 3/4 of size 1 and 1
+[16:55:56] INFO - . Diffing segment 1/4 of size 2777788 and 2777788
+[16:55:58] INFO - . Diffing segment 2/4 of size 2777788 and 2777788
+[16:55:59] INFO - . Diffing segment 3/4 of size 1 and 1
+[16:56:00] INFO - Diffing segment 1/4 of size 8333365 and 8333365
+[16:56:06] INFO - Diffing segment 2/4 of size 8333365 and 8333365
+[16:56:11] INFO - Diffing segment 3/4 of size 1 and 1
+[16:56:11] INFO - Duration: 53.51 seconds.
+```
 
 We currently support the following databases:
 
 - PostgreSQL
-
 - MySQL
-
 - Oracle
-
 - Snowflake
-
 - BigQuery
-
 - Redshift
 
+We plan to add more, including NoSQL, and even APIs like Shopify!
+
+# How to install
+
+Requires Python 3.7+ with pip.
+
+```pip install data-diff```
+
+or when you need extras like mysql and postgres
+
+```pip install "data-diff[mysql,pgsql]"```
+
+# How to use
+
+Usage: `data-diff DB1_URI TABLE1_NAME DB2_URI TABLE2_NAME [OPTIONS]`
+
+Options:
+
+  - `--help` - Show help message and exit.
+  - `-k` or `--key_column` - Name of the primary key column
+  - `-c` or `--columns` - List of names of extra columns to compare
+  - `-l` or `--limit` - Maximum number of differences to find (limits maximum bandwidth and runtime)
+  - `-s` or `--stats` - Print stats instead of a detailed diff
+  - `-d` or `--debug` - Print debug info
+  - `-v` or `--verbose` - Print extra info
+  - `--bisection-factor` - Segments per iteration. When set to 2, it performs binary search.
+  - `--bisection-threshold` - Minimal bisection threshold. i.e. maximum size of pages to diff locally.
+
 
 # How does it work?
 
@@ -63,57 +110,71 @@ We ran it with a very low bisection factor, and with the verbose flag, to demons
 
 Note: It's usually much faster to use high bisection factors, especially when there are very few changes, like in this example.
 
-```python
-$ data_diff postgres:/// Original  postgres:/// Original_1diff  -v --bisection-factor=4
-[16:55:19] INFO - Diffing tables of size 25000095 and 25000095 | segments: 4, bisection threshold: 1048576.
-[16:55:36] INFO - Diffing segment 0/4 of size 8333364 and 8333364
-[16:55:45] INFO - . Diffing segment 0/4 of size 2777787 and 2777787
-[16:55:52] INFO - . . Diffing segment 0/4 of size 925928 and 925928
-[16:55:54] INFO - . . . Diff found 2 different rows.
-+ (20000, 942013020)
-- (20000, 942013021)
-[16:55:54] INFO - . . Diffing segment 1/4 of size 925929 and 925929
-[16:55:55] INFO - . . Diffing segment 2/4 of size 925929 and 925929
-[16:55:55] INFO - . . Diffing segment 3/4 of size 1 and 1
-[16:55:56] INFO - . Diffing segment 1/4 of size 2777788 and 2777788
-[16:55:58] INFO - . Diffing segment 2/4 of size 2777788 and 2777788
-[16:55:59] INFO - . Diffing segment 3/4 of size 1 and 1
-[16:56:00] INFO - Diffing segment 1/4 of size 8333365 and 8333365
-[16:56:06] INFO - Diffing segment 2/4 of size 8333365 and 8333365
-[16:56:11] INFO - Diffing segment 3/4 of size 1 and 1
-[16:56:11] INFO - Duration: 53.51 seconds.
+## Tips for performance
+
+It's highly recommended that all involved columns are indexed.
+
+## Development Setup
+
+The development setup centers around using `docker-compose` to boot up various
+databases, and then inserting data into them.
+
+For Mac for performance of Docker, we suggest enabling in the UI:
+
+* Use new Virtualization Framework
+* Enable VirtioFS accelerated directory sharing
+
+**1. Install Data Diff**
+
+When developing/debugging, it's recommended to install dependencies and run it
+directly with `poetry` rather than go through the package.
+
+```
+brew install mysql postgresql # MacOS dependencies for C bindings
+poetry install
 ```
 
+**2. Download CSV of Testing Data**
 
-# How to install
+```shell-session
+wget https://files.grouplens.org/datasets/movielens/ml-25m.zip
+unzip ml-25m.zip -d dev/
+```
 
-Requires Python 3.7+ with pip.
+**3. Start Databases**
 
-```pip install data-diff```
+```shell-session
+docker-compose up -d mysql postgres
+```
 
-or when you need extras like mysql and postgres
+**4. Run Unit Tests**
 
-```pip install "data-diff[mysql,pgsql]"```
+```shell-session
+poetry run python3 -m unittest
+```
 
-# How to use
+**5. Seed the Database(s)**
 
-Usage: `data_diff DB1_URI TABLE1_NAME DB2_URI TABLE2_NAME [OPTIONS]`
+If you're just testing, we recommend just setting up one database (e.g.
+Postgres) to avoid incurring the long setup time repeatedly.
 
-Options:
+```shell-session
+preql -f dev/prepare_db.pql postgres://postgres:Password1@127.0.0.1:5432/postgres
+preql -f dev/prepare_db.pql mysql://mysql:Password1@127.0.0.1:3306/mysql
+preql -f dev/prepare_db.psq snowflake://<uri>
+preql -f dev/prepare_db.psq mssql://<uri>
+preql -f dev/prepare_db_bigquery.pql bigquery:///<project> # Bigquery has its own
+```
 
-  - `--help` - Show help message and exit.
-  - `-k` or `--key_column` - Name of the primary key column
-  - `-c` or `--columns` - List of names of extra columns to compare
-  - `-l` or `--limit` - Maximum number of differences to find (limits maximum bandwidth and runtime)
-  - `-s` or `--stats` - Print stats instead of a detailed diff
-  - `-d` or `--debug` - Print debug info
-  - `-v` or `--verbose` - Print extra info
-  - `--bisection-factor` - Segments per iteration. When set to 2, it performs binary search.
-  - `--bisection-threshold` - Minimal bisection threshold. i.e. maximum size of pages to diff locally.
+**6. Run data-diff against seeded database**
 
-## Tips for performance
+```bash
+poetry run python3 -m data_diff postgres://user:password@host:db Rating mysql://user:password@host:db Rating_del1 -c timestamp --stats
 
-It's highly recommended that all involved columns are indexed.
+Diff-Total: 250156 changed rows out of 25000095
+Diff-Percent: 1.0006%
+Diff-Split: +250156  -0
+```
 
 # License
 

dev/prepare_db.pql

@@ -24,7 +24,7 @@ if (db_type == "snowflake") {
     print "Uploading ratings CSV"
 
     run_sql("RM @~/ratings.csv.gz")
-    run_sql("PUT file://ml-25m/ratings.csv @~")
+    run_sql("PUT file://dev/ml-25m/ratings.csv @~")
 
     print "Loading ratings CSV"
 
@@ -86,7 +86,7 @@ if (db_type == "snowflake") {
     run_sql("create table tmp_rating(userid int, movieid int, rating float, timestamp int)")
     table tmp_rating {...}
     print "Loading ratings CSV"
-    run_sql("BULK INSERT tmp_rating from 'ml-25m/ratings.csv' with (fieldterminator = ',', rowterminator = '0x0a', FIRSTROW = 2);")
+    run_sql("BULK INSERT tmp_rating from 'dev/ml-25m/ratings.csv' with (fieldterminator = ',', rowterminator = '0x0a', FIRSTROW = 2);")
     print "Populating actual table"
     rating += tmp_rating
     commit()
@@ -99,7 +99,7 @@ if (db_type == "snowflake") {
         rating: float
         timestamp: int
     }
-    import_csv(rating, 'ml-25m/ratings.csv', true)
+    import_csv(rating, 'dev/ml-25m/ratings.csv', true)
     rating.add_index("id", true)
     rating.add_index("timestamp")
     run_sql("CREATE INDEX index_rating_id_timestamp ON rating (id, timestamp)")

dev/docker-compose.yml → docker-compose.yml

@@ -1,44 +1,12 @@
 version: "3.8"
 
 services:
-    data-diff:
-      container_name: data-diff
-      build: 
-        context: ../
-        dockerfile: ./dev/Dockerfile
-        args:
-          - DB1_URI
-          - TABLE1_NAME
-          - DB2_URI
-          - TABLE2_NAME
-          - OPTIONS
-      ports:
-        - '9992:9992'
-      expose:
-        - '9992'
-      tty: true
-      networks:
-        - local
-
-    prepdb:
-      container_name: prepdb
-      build: 
-        context: ../
-        dockerfile: ./dev/Dockerfile
-      command: ["bash", "./dev/prepdb.sh"]
-      volumes: 
-        - prepdb-data:/app:delegated
-      ports:
-        - '9991:9991'
-      expose:
-        - '9991'
-      tty: true
-      networks:
-        - local
-
     postgres:
       container_name: postgresql
       image: postgres:14.1-alpine
+      # work_mem: less tmp files
+      # maintenance_work_mem: improve table-level op perf
+      # max_wal_size: allow more time before merging to heap
       command: >
         -c work_mem=1GB
         -c maintenance_work_mem=1GB
@@ -51,25 +19,22 @@ services:
       expose:
         - '5432'
       env_file:
-        - dev.env
+        - dev/dev.env
       tty: true
       networks:
         - local
 
     mysql:
       container_name: mysql
       image: mysql:oracle
+      # fsync less aggressively for insertion perf for test setup
       command: >
         --default-authentication-plugin=mysql_native_password
-        --innodb-buffer-pool-size=8G
-        --innodb_io_capacity=2000
-        --innodb_log_file_size=1G
         --binlog-cache-size=16M
         --key_buffer_size=0
         --max_connections=10
         --innodb_flush_log_at_trx_commit=2
         --innodb_flush_log_at_timeout=10
-        --innodb_flush_method=O_DSYNC
         --innodb_log_compressed_pages=OFF
         --sync_binlog=0
       restart: always
@@ -81,16 +46,15 @@ services:
       expose:
         - '3306'
       env_file:
-        - dev.env
+        - dev/dev.env
       tty: true
       networks:
         - local
 
 volumes:
   postgresql-data:
   mysql-data:
-  prepdb-data:
 
 networks:
   local:
-    driver: bridge
+    driver: bridge