Distinguish between actual content edits and reactions in edited_message updates

[sgdc3]

Currently, when a message receives a reaction, bots can receive an edited_message update even if the message content (text, media, or attachments) has not actually been modified.

This makes it difficult, if not impossible, to understand whether the update is due to an actual change in content or simply an added/removed reaction.

Proposal

It would be very useful to have a dedicated field in the edited_message payload that clearly indicates whether the content has changed.
For example:

“content_changed”: true // or false

If it is not possible to add this field, any other way to clearly distinguish between content changes and reaction changes alone would still be welcome.

Why it matters

• It avoids having to rework or resend messages unnecessarily when only a reaction has changed.
• It allows the development of bots that monitor possible abuse of changes to old messages (e.g., media replacement) by immediately distinguishing real changes from simple reactions.

Related issues

Probably related to #388.

[RememberTheAir]

Since there's a set of methods dedicated to receiving reactions updates, I consider this a bug in the bot api. It is unexpected to receive such updates for messages which content has not been modified according to what "editing messages" means on telegram.

To workaround this issue I store the last received text, caption and unique_file_id for every message update received, and every time and edited_message update is received I check whether these three fields are unchanged for that message. In the related issue you mention, which describes the same problem, it looks like this isn't going to be fixed

[AYMENJD]

@RememberTheAir, edited_message is sent to inform the bot about the up-to-date message state, and duplication should not cause any harm.

Even before reactions existed, edited_message was never meant to only signal "content" changes. There are cases where you may receive an edited_message update with no visible changes.

Fixing this won’t help, since Telegram can introduce another feature at any time that breaks the assumption that "edited_message is only for content changes."

[levlam]

If it is not possible to add this field, any other way to clearly distinguish between content changes and reaction changes alone would still be welcome.

"edited_message" updates notify bots that some data about a message could have been changed. The update is not limited to "content". or "reactions", or whatever else changed. Even the content has changed, the change may be irrelevant for the bot like presence of spoiler, position of caption, video start timestamp, inline buttons, or anything added to the message content in the future like suggested post information added in the recent update. Any abstraction like "content change" is just a bug in the bot's code. If the bot is supposed to check whether message text and media are the same, then it must compare them and only them.

Bots could have ever received "edited_message" updates, while nothing visible for the bot was changed. For example, such updates are received when a web page is added to the message, or the message is automatically forwarded to discussion group, or the message is restricted for some users (for example, for copyright violation). There were many such cases since ever and there will be much more in the future.

[sgdc3]

Thanks for the clarifications, I get why edited_message is a “message state changed” signal rather than “content changed.”
That said, I’m still trying to find a practical way for a bot to detect user-visible text/media changes without keeping a heavy database for every message in large groups.

If the bot is supposed to check whether message text and media are the same, then it must compare them and only them.

The only robust approach I see is fingerprinting each message (hashing text/caption + media identifiers) and comparing on every
edited_message. At scale, though, storing per-message state across many big groups gets expensive.
@levlam Before I dive deeper into that approach, could you confirm whether there’s a simpler way to detect content changes using the existing API features?

[levlam]

Before I dive deeper into that approach, could you confirm whether there’s a simpler way to detect content changes using the existing API features?

There is no.

At scale, though, storing per-message state across many big groups gets expensive.

Messages can be edited only for 2 days by regular group members. I see nothing expensive in storing hashes of all messages for 2 days.

[RememberTheAir]

Messages can be edited only for 2 days by regular group members. I see nothing expensive in storing hashes of all messages for 2 days.

But channel messages and messages by admins with the permission to edit messages can be edited with no time limit. As far as I remember, once an user is promoted to admin and given the "pin messages" permission, they can edit even the messages that were sent when they were regular users - so one would have to store indefinitely any channel/supergroup message (correct me if I am wrong).

Btw I am using the text hash/caption hash/file_unique_id triplet to detect content changes and it is reliable and not too much of an overhead on my bots' operations. It took me some time debugging this the first time I came across these "weird" updates but that's because I didn't carefully read the docs

[levlam]

But channel messages and messages by admins with the permission to edit messages can be edited with no time limit.

Yes, but I highly doubt that the bot is interested in edits of messages by administrators of the chat:

It allows the development of bots that monitor possible abuse of changes to old messages (e.g., media replacement)

[sgdc3]

The suggested approach works correctly, I am closing the issue.