Pitfalls of defillama token data

[0xngmi]

General

• We used to get token symbols from on-chain queries, but some tokens had non-standard symbol() methods so we couldn't get their symbols properly, due to this we switched over to getting the symbols from coingecko. Token symbols might also change mid-way through the data due to this if coingecko's symbol is different from the on-chain one. When we were getting symbols from on-chain we labelled symbols for contracts that were broken as UNKNOWN(address).
• We rely on coingecko queries by address for token prices and there's plenty of cases where coingecko doesn't have the address of some token or just doesn't list the token. In these cases we replace the token with one that is equivalent (eg: we'll replace the USDJ for USDT since they are both stablecoins, or we'll replace renLUNA for LUNA).
• ethereum and WETH are the same, initially we used ethereum for native ETH and WETH for WETH, but later we transitioned into grouping everything under WETH. On the /protocol endpoint we normalize it by adding them together under WETH but on other data sources such as the csv we don't apply that normalization.
• If coingecko doesn't have a token that is part of TVL but contributes too little to TVL we just don't count it. This normally doesn't happen on most projects but we've applied it to projects that have a lot of tokens under TVL such as pancakeswap. The difference in TVL when counting these tokens is minimal.

Projects with no tokens

You'll see that there's some projects for which tokens are just missing from the API, this ocurs for projects in which we get their TVL directly from an API which returns the data in USD, so there's no point in storing tokens for them. Most projects don't fall into this category, the main ones that fall here are yield aggregators.

Projects with wrong tokens

There are some projects for which we deliberately misrepresent their tokens. This happens in projects such as pancakeswap, where we get their TVL from a subgraph and we report all of it as USDT.
You can find these projects by looking for the property misrepresentedTokens on the /protocol/{slug} endpoint

Fixes

If you notice something wrong, a project that has misrepresented tokens or one that is missing them, please add them to a list. Re-writing adapters to get tokens is not a priority (although I regularly re-write them to get historical data or improve our transparency and then I fix the token reporting) but a list would be nice for when we people ask us what they should contribute.