Setting up ADAL/MSAL for connecting to Azure database with Matlab

After further inquiries with MathWorks I was able to resolve my issue. The main problem was that I was missing the ADAL library and all its dependencies. After installing them, I was able to connect to the database. In the end, this was the setup that got everything working.

• Java: At first we installed the AdoptOpenJDK Java, but that was a mistake and not necessary
  • What I did: I rolled back to the native Matlab Java. I removed all references to the new Java in the Environmental Variables such that Matlab could only use the default Java that it is shipped with.
• JDBC driver: Is it strictly necessary to use the JDBC driver, or could you use an ODBC connection as well? I don't exactly know the answer to this question. In Matlab it seems that if you want to programmatically, via a script, connect to the database, you need/want to use a JDBC driver. Although this was not explicitly stated by the MathWorks support staff, they did not suggest to use ODBC either
  • What I did: Use the latest JDBC driver that still supports the default Java version of Matlab, Microsoft JDBC DRIVER 6.4 for SQL Server for Matlab 2017a (it still uses Java 7) and Microsoft JDBC DRIVER 8.4 for SQL Server for Matlab 2020a (it uses Java 8)
• ADAL libraries: Matlab requires the ADAL4J library and all of its dependencies. From this I conclude that using MSAL4J is not a possibility. I got the necessary libraries directly from MathWork, so I have no clue on how to obtain them. The libraries I used are:

accessors-smart-1.2.jar
activation-1.1.jar
adal4j-1.6.3.jar
asm-5.0.4.jar
commons-codec-1.11.jar
commons-lang3-3.5.jar
gson-2.8.0.jar
javax.mail-1.6.1.jar
javax.servlet-api-4.0.1.jar
jcip-annotations-1.0-1.jar
json-smart-2.3.jar
lang-tag-1.5.jar
nimbus-jose-jwt-9.0.1.jar
oauth2-oidc-sdk-5.64.4.jar
slf4j-api-1.7.21.jar

This concludes the Java/driver related parts of the answer. To setup everything in Matlab the following steps are necessary.

1. Matlab needs to know where the JDBC driver is installed and where the ADAL4J libraries can be found
2. Each Matlab session, you need to override a Matlab setting to avoid that Matlab forces third-party Java classes to use an older Saxon Transformer. This older Saxon Transformer is not compatible with ADAL4J.

Solution:

1. In Matlab execute the command edit(fullfile(prefdir,'javaclasspath.txt')). If the file does not exists yet, you will be prompted to create the file. Allow Matlab to create the file in that case.
2. In the .txt file, enter the following lines:

   <before>
   c:\full\path\to\accessors-smart-1.2.jar
   c:\full\path\to\activation-1.1.jar
   c:\full\path\to\adal4j-1.6.3.jar
   c:\full\path\to\asm-5.0.4.jar
   c:\full\path\to\commons-codec-1.11.jar
   c:\full\path\to\commons-lang3-3.5.jar
   c:\full\path\to\gson-2.8.0.jar
   c:\full\path\to\javax.mail-1.6.1.jar
   c:\full\path\to\javax.servlet-api-4.0.1.jar
   c:\full\path\to\jcip-annotations-1.0-1.jar
   c:\full\path\to\json-smart-2.3.jar
   c:\full\path\to\lang-tag-1.5.jar
   c:\full\path\to\nimbus-jose-jwt-9.0.1.jar
   c:\full\path\to\oauth2-oidc-sdk-5.64.4.jar
   c:\full\path\to\slf4j-api-1.7.21.jar
   c:\full\path\to\<SQL JDBC driver>.jre<relevant number for the java version you use>.jar
   

   • The <before> tag will ensure that the libraries are added to the front of the Static Java Path in Matlab.
   • Replace c:\full\path\to\ with the directory where Matlab can find the JDBC driver and the ADAL4J libraries.
   • For the JDBC driver, select the .jar files that corresponds to the Java version that your Matlab uses.
3. Save javaclasspath.txt and restart Matlab. Use the command javaclasspathin the command window to verify that the libraries are added to the Static Java Path
4. Steps 1-3 need to be repeated per user per Matlab release. It is also possible however to only set it up once. Instead of using javaclasspath.txt in the preference directory of the user, you can also update classpath.txt in the toolbox\local directory of the Matlab installation. If you choose to update classpath.txt, just add the these lines to the top off the file and do not use the <before> keyword.
5. Each Matlab session, i.e. each time you start up Matlab (yes, really, each time), before connecting to the database, execute the following command:

   java.lang.System.clearProperty('javax.xml.transform.TransformerFactory')
   

   This reverts the java.xml.transform.TransformerFactory to its default Java setting. MATLAB will normally have overridden this setting which would make third-party Java classes work with an older Saxon Transformer, this is incompatible with ADAL4J though. So, we reset this to the default such that ADAL4J can use the default Java XML Transformer again. This has to be done once per MATLAB session.
6. Lastly you can use the following command to connect to your database:

   conn = database('myDatabase','user@domain.com','myPassword',...
   'com.microsoft.sqlserver.jdbc.SQLServerDriver',...
   ['jdbc:sqlserver://myServer.database.windows.net:1433;' ...
    'encrypt=true;trustServerCertificate=false;' ...
    'hostNameInCertificate=*.database.windows.net;' ...
    'loginTimeout=30;authentication=ActiveDirectoryPassword;database='])
   

   Where you will need to set the actual database name, username, password and server address.

That's it, you should now be able to connect to your database!