Special commands inside `source` inside `--execute` are not supported, but the error is not handled correctly.

[your-diary]

Setup

$ mycli --execute 'source 1.sql' test_database

where 1.sql contains a special command (source, \T, etc.):

SELECT 1;
\T ascii; -- In addition to `\T`, I also confirmed with a nested `source`
SELECT 2;

Expected Behavior

• An error message is printed.

• The command returns a non-zero exit status.

Actual Behavior

• Processing the file is aborted right after encountering a special command (this is preferable).

  • In other words, SELECT 1 is executed while SELECT 2 is not executed.

• No error message is printed.

• The command returns 0 as the exit status.

System

• mycli version: v2.13.2
• OS/version: Alpine Linux

Cause (AI-generated, unverified)

The problem is that the rejection is returned as an SQLResult.status.
The --execute path does not render that status and treats the completed
run_query() call as success, causing silent partial execution with exit
status 0.

[rolandwalker]

Hi! Great report. We kinda knew everything around source was not perfect, but specifics are very helpful.

Why do --execute 'source 1.sql' instead of --batch 1.sql? Would it be acceptable to disallow source inside of --execute? Because I'd be inclined that way.

[your-diary]

Hi, thank you for the reply.

Would it be acceptable to disallow source inside of --execute?

No.
As I embed mycli --execute into my own program executor Vim plugin, if possible, I want source inside --execute to continue to be supported (even if --batch could technically replace it).

Why do --execute 'source 1.sql' instead of --batch 1.sql?

One reason is that, when I created the plugin, I couldn't find a (at least an easy) way to change the output format dynamically (e.g. via \T mysql_unicode).

Another reason is customizability. --execute can be extended just by modifying the command line argument like --execute 'SET some_system_variable = ...; \T some_format; ...; source 1.sql;' while I have to modify the file 1.sql for --batch, which is not always feasible because I often bind-mount SQL files inside a Docker container read-only (or maybe I have to write a more complex pipeline to utilize --batch - (e.g. cat <file> | append_string ... | mycli --batch -).
I know there is --init-command, but I believe mycli behaves (a bit?) differently between --init-command and --execute.

---

By the way, just in case, please note the issue is not proposing to add support for special commands inside source inside --execute; better error handling is enough.

Also, for me, "close as not planned" is better than dropping support for source inside --execute.