Hashes differ from ones on pinata.cloud

[aradica]

They don't start with Qm...?

[saeta-eth]

There are two versions of CIDs. Have you seen them? I think it's related to your issue.

Related links:

• https://docs.ipfs.io/concepts/content-addressing/#version-0-v0
• https://docs.ipfs.io/concepts/content-addressing/#version-1-v1

[fandrupaw]

hey, Im facing same issue, Pinata by default uses CID v0, both hashes starts with Qm, the rest is different

[alanshaw]

@aradica see the comment here https://github.com/alanshaw/ipfs-only-hash/issues/20#issuecomment-983493673 TL;DR you can pass --cid-version=0 to get Qm... hashes on the command line.

@fandrupaw any chance you can share the data, and the CID you're getting from Pinata so we can verify? 🙏 The unixfs importer code is interop tested with go-ipfs to ensure the same CID is produced for the same data but there could be a bug!

[titusz]

ipfs-only-hash does also seem to create different CIDv1s when compared to ipfs-go version 0.11.0: Testing for example the bitcoin whitepaper

ipfs add --only-hash --cid-version=1 bitcoin.pdf
added bafkreifrm5azdkeoyxg5om7eeqfidabraxoecllmm4enkovzj7ber5hvkm bitcoin.pdf

versus

ipfs-only-hash --cid-version=1 bitcoin.pdf
bafybeibj3nf4iyxt2guxihs77sylpuwu4l4yn4cfqumpc2xplxgxt4ssoa

[MAkcanca]

From here and checking the two CIDs you provided, it's obvious that one of them is wrapped with UnixFS, and other one is raw block. That's why the CIDs are different

[chichke]

Same here works great with CIDv0 but doesn't with CIDv1

[naviangie]

The same chunking and encoding algorithm is required.

importer([{content}], block, { onlyHash: true, cidVersion: 1, rawLeaves: true })

[larskarbo]

I had some issues getting the same hashes as Pinata for files above 262kb (when chunking kicks in).

This is my solution using Helia:

import { unixfs } from "@helia/unixfs"
import { BlackHoleBlockstore } from "blockstore-core/black-hole"
import { fixedSize } from "ipfs-unixfs-importer/chunker"
import { balanced } from "ipfs-unixfs-importer/layout"

export const calculateCid = async (bytes: Uint8Array) => {
    const unixFs = unixfs({
      blockstore: new BlackHoleBlockstore(),
    })

    const cid = await unixFs.addBytes(bytes, {
      cidVersion: 0,
      rawLeaves: false,
      leafType: "file",
      layout: balanced({
        maxChildrenPerNode: 174,
      }),
      chunker: fixedSize({
        chunkSize: 262144,
      }),
    })

    const cidv0 = cid.toV0().toString() // QmPK1s...
    const cidv1 = cid.toV1().toString() // b45165...
}

I wrote some more about it here.

[lancelot-c]

Finally something that works ! Thanks @larskarbo !

[endomorphosis]

I also noticed that there are differences in hashes between ipfs command and this package.

['bafybeif7ztnhq65lumvvtr4ekcwd2ifwgm3awq4zfr3srh462rwyinlb4y'] -- this package ['bafkreihdwdcefgh4dqkjv67uzcmw7ojee6xedzdetojuzjevtenxquvyku'] multiformats python library ['bafkreihdwdcefgh4dqkjv67uzcmw7ojee6xedzdetojuzjevtenxquvyku'] ipfs kubo