Visual Studio Code - Development Container - PostgreSQL
This project provides a Visual Studio Code (VSCode) development container for PostgreSQL. This container contains all the tools needed to debug PostgreSQL and a PostgreSQL build that is compiled without optimizations, enabled assertions, and debug symbols. This PostgreSQL build is slower than a production build but allows it to develop and test PostgreSQL extensions. For instance, the build performs checks to catch programming mistakes due to the enabled assertions (USE_ASSERT_CHECKING) and cleaning freed memory early (CLOBBER_FREED_MEMORY). The created debug symbols and the missing optimizations make it easy to attach a debugger to this instance.
Prerequisites
To use the development container, the following prerequisites have to be met:
Open the Project
After the prerequisites are met, the VSCode project can be opened. After loading the project, VSCode asks if the project should be Reopened in Container; this option should be selected.
Afterward, VSCode downloads the needed base container image, installs the tools, and compiles PostgreSQL. This step might need some time to complete. Details about the build process are shown by clicking on the show log link.
First Start
After the container is built, the PostgreSQL source code that is part of the development container should be added to VSCode. This allows you to browse the source code and perform full-text search from VSCode as well. To import the source code, a new terminal should be opened and the following command should be executed:
code --add /usr/local/src/postgresql
Note: VSCode might need to reload the window after executing the command.
For example, you can search for the PostgreSQL function add_path() after the header files (.h) of the source code is added to VSCode, and you should find some matches.
Connecting to the PostgreSQL server
The development container automatically starts the PostgreSQL server when the container is executed. Using the integrated terminal, you can connect to this PostgreSQL server after you have created a database using createdb. For example, the following commands create a database, connect to the database, create a table, perform a SELECT statement, and delete the database.
createdb test
psql test
CREATE TABLE test (id INTEGER);
SELECT * FROM test;
\q
dropdb test
Installing an Extension
The container also contains a simple hello_world PostgreSQL extension, which provides the function hello_world(TEXT). The function takes a parameter of the type TEXT and returns Hello %PARAMETER also as a TEXT. The source code of the extension can be found in the src/extension/01_hello_world/ directory.
To install and test the extension, you need to change to the directory of the extension. Afterward, you can compile it by running make. This builds the library hello_world.so. By running sudo make install, the library is copied into the lib folder of the database server, and helper files are copied into the extensions directory of PostgreSQL. The command must be prefixed with sudo since only root has write permissions on these directories. After the extension is installed it can be tested. By running make installcheck a new database is created and, the test statements of the file test/sql/00_hello.sql are executed and the results are compared with the output stored in test/expected/00_hello.out.
cd src/extensions/01_hello_world/
make
sudo make install
make installcheck
The extension can be also tested interactively. To use the extension, a new database needs to be created first:
createdb test
psql test
Afterward, the extension should be available for installation. This can be verified by the following command:
SELECT * FROM pg_available_extensions;
name | default_version | installed_version | comment
--------------+-----------------+-------------------+----------------------------------------------------
postgres_fdw | 1.1 | | foreign-data wrapper for remote PostgreSQL servers
plpgsql | 1.0 | 1.0 | PL/pgSQL procedural language
hello_world | 1.0 | | Simple hello world extension
(3 rows)
It shows that the extensions postgres_fdw, plpgsql, and hello_world are known by the database server.
By executing:
CREATE EXTENSION hello_world;
the extension can be installed in the current database. By executing \dx (a helper command provided by psql), the installed extensions are listed:
\dx
List of installed extensions
Name | Version | Schema | Description
-------------+---------+------------+------------------------------
hello_world | 1.0 | public | Simple hello world extension
plpgsql | 1.0 | pg_catalog | PL/pgSQL procedural language
(2 rows)
During the installation, the content of the install script of the extension hello_world--1.0.sql is executed. The script registers a new function in the database catalog and defines the entry point in the C code. In other words, the function is known at the SQL level afterward and PostgreSQL knows that it should call the implementation in our extension. By running the helper command \df+ hello_world the information about the function can be retrieved from the catalog.
\df+ hello_world
Schema | Name | Result data type | Argument data types | Type | Volatility | Parallel | Owner | Security | Access privileges | Language | Internal name | Description
--------+-------------+------------------+---------------------+------+------------+----------+--------+----------+-------------------+----------+---------------+-------------
public | hello_world | text | text | func | volatile | unsafe | vscode | invoker | | c | hello_world |
The function can now be called:
SELECT hello_world('Mr X');
hello_world
-------------
Hello Mr X
(1 row)
Using a Debugger
To debug the extension, the source code hello_world.c needs to be opened. Afterward, a breakpoint can be set by clicking on the dot right of the line number.
To stop the execution at this point, open a new connection to the database server by running psql test from the terminal.
Afterward, click on the editor and press F5. A drop-down menu is shown, which asks to which process the debugger should connect. Type test and you should see two PostgreSQL processes containing the name of the test database. One process is the psql client, the other process is the actual PostgreSQL server backend. It is named postgresql: vscode test [local] idle. vscode is the user that is connected to the database, test is the database name, [local] is the name of the client host, and idle is the transaction status of this backend. This process should be selected from the drop-down list. Afterward the debugger attaches to this process.
Afterward, execute the command SELECT hello_world('Mr X'); again from your terminal (there might be multiple terminals now, one for the debugger and your terminal executing psql. Make sure you select the right one from the list on the right in the terminal tab).
After the command is executed, the debugger stops at the breakpoint. On the left pane, you see the actual values of the Locals (the variables that are visible in the scope). In addition, the call stack is shown below, and you can switch between the frames on the call stack. On the top, you find some new buttons that allow you to continue the execution or perform an execution line by line or enter the called functions and execute them line by line.
Note: You can also switch to the implementation of the used functions by pressing F12. For example, click on the function text_to_cstring() and press F12 afterward. The file varlena.c of the PostgreSQL source code is opened in a new tab, and the cursor is moved to the function definition.