Fixed the issue with last character handling for cases when the length of the message is 4001 characters

[maxtrap]

Pull Request Etiquette

• I have checked the PRs for upcoming features/bug fixes.
• I have read the contributing guidelines.

Changes

• Internal code
• Library interface (affecting end-user code)
• Documentation
• Other: _____

Closes Issue: NaN

Description

This fixes the issue that happens when MessageBuilder#buildAll is called on a message with 2001+ characters. Once the message is split, the last character of the last message ends up being cut off. The reason this happens is because the endIndex of the substring is set as builder.length()-1, even though a substring's end index is exclusive. As a result, the last char in the StringBuilder of the message ends up being cut off.

Tested

With the following sample code

        StringBuilder db = new StringBuilder();
        for (int i = 0; i < 4000; i++) {
            if (i != 1999 && i != 3999)
                db.append('a');
            else db.append('b');
        }
        db.append('c');

        new MessageBuilder(db.toString())
                .buildAll(MessageBuilder.SplitPolicy.ANYWHERE)
                .forEach(replyMessage -> getServer().getTextChannelById(CHANNEL_ID).sendMessage(replyMessage.getContentRaw()).queue());

Existing implementation loses last 'c' letter, by generating 2 messages

And proposed implementation correctly generates 3 messages

[DManstrator]

Good catch!

Small suggestion: Use the constant from the Message interface (Message.MAX_CONTENT_LENGTH) instead of the 2000.

Edit: Same for line 1277.

[Almighty-Alpaca]

I agree with the comment from @DManstrator but otherwise lgtm

[maxtrap]

Thanks for review.
Addressed comment by replacing hardcoded number with constant.
Please take a look again.

[DManstrator]

otherwise lgtm

lgtm means "looks good to me". So it is fine now. 👍