kaspa-live / kaspa-graph-inspector Goto Github PK

After leaving kaspa.live open for some time, I've noticed that no more blocks are being added.

In the console there was a single error message:

Uncaught (in promise) TypeError: e.mergeSetRedIds is null
    M Block.ts:19
    V BlocksAndEdgesAndHeightGroups.ts:24
    setBlocksAndEdgesAndHeightGroups TimelineContainer.ts:65
    e Dag.ts:247
    c runtime.js:63
    _invoke runtime.js:293
    E runtime.js:118

Clicking a block immediately revealed the rest of the DAG, and the >> button.
Clicking the >> button resumed operation to normal.

Mempool should be limited to 1mil transactions, which translates to about 300 megabytes.
However, we saw a node in which (seemingly) mempool transactions took up to 2.3 gigs.

Write a stability-test that:

1. Fills the mempool up to the limit
2. Keeps adding more transactions, making sure eviction works
3. Starts mining blocks to make sure this cleans the mempool eventually.

Sometimes I'd like to illustrate what's the red block to people, and since it's a rara avis I have to scroll back the KGI for many minutes to find one. Would be cool to have the way to simply click the "take me to the next red block in the past from where I am now" option

There's too much free space to the right of the tip of the DAG in some modes of the browser window: this is what it looks like on a FullHD monitor when the Firefox browser is in the fullscreen mode

and this is how it looks when the browser is not maximized:

Too much free space to the right in the first case.

The bug is as follows:

• There's an edge between a block at height 200 and a block at height 100
• The screen may contain up to 50 heights
• The target height is 150, meaning that only blocks in heights 125 to 175 are shown
• The edge is not shown

"And tell me how far back it is from the current one of the kind" Could be a nice instrument for "manually" investigating the DAG statistics and proving its params for being optimal or not :)

P.S. Also the same "how far from the current one of the kind" could be added to the red blocks search instrument as well. Probably just a momentarily floating hint.

P.P.S. Also then it's better to have both "search back" and "search ahead" options for both functionalities :) since I could be anywhere in the graph at a given time.

As discussed in discord (see https://discord.com/channels/599153230659846165/951517014873235497/951518516056580179), some instructions in the readme would be very welcomed in order to deploy a mirror of the main web site.

It can be useful to be able to clearly see the the anticone of a block, and in the case of a red block, the blocks that prevent it from being blue i.e. the blocks in its anticone which already have k blue blocks in their respective anticone.

For both of the above, as well as merge set, a clear visualization could be nice, where you could highlight the respective blocks in some way.

Example: http://kgi-testnet-3.daglabs-dev.com/?hash=ecc8bd82b1fbc5e63f270d081a790a7a5b1adbecaf9d458ef6c3689820682186

Proof that this is a real block: http://testnet-3.katnip.sh/#/block/ecc8bd82b1fbc5e63f270d081a790a7a5b1adbecaf9d458ef6c3689820682186

Sometimes i'd like to see a wider picture of the DAG 🙂 And that requires me shrinking the browser vertically. It would be much more convenient to have the ability to make it by hands inside the site.

A 'testnet' message on the testnet kgi would make clear for users that they are not in the mainnet.

There are times when we have much more blocks in parallel than it is possible to fit onto screen vertically. Can you maybe add a switch there that will turn on "compressed" visualization of such blocks? So they would be either overlaying partially, or maybe have the not a square view, but a rectangular one. So that all of these many vertical blocks fit onto screen at once.

This is because there's essentially zero error handling.

Proposed solution:

• Catch and count subsequent errors
• If the error count is greater than some N, indicate that there's a problem somewhere on screen
• Clicking the indication should hard-refresh the page

Every so often, there are virtual selected parent chain edges that never get marked (bold and blue)

Collapsing the Dag to the consensus topological ordering, or even allowing for a dual view, could be very useful for both analysis and educational purposes.

My team is going to write an app very similar to the kaspa-graph-indicator and was wondering if we could use the frontend part of the repo as a framework. We'd probably modify a bunch of stuff to make it work. The repo is public however it has no license so we cant really use it since it follows the default copyright rights. So my question is if some license could be added?

For reference info on repos without a license

This happens as follows:

• A block is added to some height
• Later, another block is added to the same height
• The old block "pops" into its new y position