shuhaowu opened a new issue, #84:
URL: https://github.com/apache/arrow-js/issues/84
### Describe the enhancement requested
Certain codebases that previously uses row-oriented way to access data may
wish to migrate to Arrow to save serialization and deserialization cost, and to
be able to gain access to fast column-oriented operations. As it stands, Arrow
is sort of a drop-in replacement to row-oriented data such as a JavaScript
Array of objects. This is great to incrementally migrate legacy codebases to
Arrow, as it is frequently infeasible to rewrite the application to use the
column-oriented data access patterns. For most data,
JavaScript-object-compatible and row-oriented access is already provided via
the `StructRowProxy`. However, if the structs themselves include a `Vector`,
existing code will break as it assumes the `Vector` object to behave like a
JavaScript array, which it does not due to the lack of index access. An example
of such a data structure is as follows:
```
[
{x: 1, y: [1, 2]},
{x: 2, y: [2, 3]},
]
```
In this case, with the Arrow JS library as it is, the API consumer is unable
to get individual element of the `y` array via `table[i].y[j]`. Instead, the
API consumer must use the API `table.get(i).y.get(j)`. In the situation where
we are migrating a legacy code base to Arrow, this requires a large refactor of
the entire codebase, which is infeasible in a short time. This negates the
advantage of using Arrow as a drop-in replacement and prevents incremental
migration of code to Arrow.
Arrow should provide index access for the `Vector` object, as well as
`Table` and `RecordBatch` for backward compatibility with JavaScript arrays.
### Component(s)
JavaScript
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
