4 ms·
Two issues: First, to be a modeling language, tools have to be able to discover entities and relationships by reading a model definition. Second, to be successf
by petilon 5y ago
Two issues: First, to be a modeling language, tools have to be able to discover entities and relationships by reading a model definition. Second, to be successful as a modeling language, users need to be able to author models easily. If it has to be hand-coded it will not be successful.
For the first issue, this language has a non-standard syntax. It doesn't use JSON or XML. So either I have to write a parser on my own, or Google has to provide parsers in every language. Also, models are expressed as "explores" [1] where they define the fields available, and the query to use to get the data. But that's not what I expect from a modeling language. The purpose of modeling languages is to enable users to write new ad-hoc queries on the fly. If the queries have to be pre-written into an "explore" then this is not really a modeling language, it is just a query language.
I don't see this taking off as a modeling language.
[1] https://looker-open-source.github.io/malloy/documentation/language/explore.html https://looker-open-source.github.io/malloy/documentation/la...