four_bytes

README

Four-bytes database generator

make
./bin/four_bytes --outdir /where/to/put/chunks

It generates the cross-product database of four-bytes from the data already present in TrueBlocks. The input files are installed with chifra:

• abis/known-000/uniq_sigs.tab (Set A)
• abis/known-000/uniq_funcs.tab (Set B)
• abis/known-000/uniq_events.tab (Set B)

Set A has 784 items.

Set B has 113,281 items in total.

The produced database is chunked and consists of 65,536 files, with sizes ranging from 100 KB to 176 KB. The database takes about 7 GB of disk space in total. Generation can take up to 40 minutes.

Chunk format

Chunk format tries to follow Unchained Index file format as close as possible. Each chunk consist of 3 parts:

[      HEADER      ]
[ Signature Table  ]
[   String Table   ]

Header

type Header struct {
	Magic          uint32
	Hash           common.Hash
	SignatureCount uint32
}

Signature Table

is array of

type SignatureRecord struct {
	Signature common.Hash
	// Where the string representation starts in the file?
	Offset    uint32
	// How long is the string representation?
	Len       uint32
}

String Table

is strings packed next to each other, e.g.: islets()tixFounders()reclaimToken(address)partisans(uint256)setPurchaseValues(uint128)

Further improvements

1. Jay:

   This might speed things up very significantly and not lose very much data. A while ago, I looked at about 850,000 functions (I forget where I got them). I extracted the type signatures. There were about 45,000 uniq type signatures, but the attached 500 accounted for 99% of the entries.

2. Cross-product database has many entries that has been never seen on chain. Maybe we could add frequency information to give priority to the real signatures