normharrison / ada_buffered_streams

A novice implementation of a buffer around Ada's built-in stream type.

Ada buffered streams

A personal WIP project meant to hopefully improve the performance of stream-based communication in other projects. Buffered writing and reading is provided, with reading handled in two different ways depending on use case.

The root type is declared as abstract inside buffered_streams.ads and provides the implementation of buffered writing that applies to all descendant types. Buffered writing works by directing all calls to Write to an internal buffer, then sending the buffer's contents across the stream all at once when it's completely filled, or whenever the Flush primitive is explicitly called.

The first implementation of buffered reading is provided by the subtype Agnostic_Buffer_Type inside the package buffered_streams-agnostic_buffer.ads. This type works much like a regular stream, in which any data type can be read from the stream at any time. Buffering functionality works by calling the Prefill primitive and specifying the size of the type that will be read next. Calls to Prefill are currently always expected to be followed by calls to Read. The way in which buffering works via this type is planned to change in the future.

The second implementation of buffered reading is provided by the subtype Unique_Buffer_Type inside the package buffered_streams-unique_buffer.ads. This type works differently than traditional streams, as indicated by its name, it is meant to be used to only read a single data type from the stream, like characters/strings for example, hence why the package is generic. Through this restriction, it is able to offer the ability to read until a certain delimiter is encountered in the buffer via the Read_Until primitive, returning all data that was read up until that point on Ada's secondary stack. Though in some cases you may need to explicitly allocate on the heap when very large amounts of data are being dealt with.