Shards Lifecycle¶
When you run a Shards script, it goes through specific stages before the program is executed. You don’t need to know these stages to write Shards scripts, but this page explains each stage to help you understand what happens under the hood of Shards.
The Lifecycle¶
Read¶
The Shards text .shs file is read.
Parse¶
The read .shs file is converted to AST format.
.sho file
shards build can be run to convert your .shs file into an AST .sho file.
Construct¶
The Shards computational graph is constructed from the parsed AST file.
What is a Shards computational graph?
A computational graph is an intermediate representation of your programme meant for execution. While a Shards text .shs and even an AST .sho file is formatted for human reading and understanding, a computational graph is specifically meant for computers to understand and process quickly.
#
When you prefix a parenthesized expression with #, eg. #(3 | Math.Add(2)), it is evaluated at this stage and the resulting value is embedded.
Compose¶
The constructed computational graph is composed. Here, each output, input and parameter are validated (checked if they are the correct type for example). Optimizations and the fixing of output types also happens at this stage.
Warmup¶
At this stage, objects are created, memory allocated or pre-allocated and data is prepared.
Shards efficiency
The Warmup stage is unique to Shards as a coding language and is one main reason why shards is so efficient at running programs. The Warmup stage happens only once per wire lifetime. So when a looped wire is scheduled, warmup happens once, then at the activation stage, the looped wire then only has to reuse all the resources that has been prepared at the warmup stage each time it is activated. Consequently, when a non-looped wire is activated multiple times, warmup → activation → cleanup happens at every activation.
Activation¶
The actual running of our program happens at this stage. Ideally, only minimal work happens here because of Warmup.
Cleanup¶
Memory is freed and deallocated, and resources cleaned up.