Recipes in two minutes

What a recipe is, what it consists of, and why there are no paths or strings around it.

A recipe is a named block of actions. It can depend on other recipes and use their results. Users write their own recipes; there are no built-in “system recipes”. What is built in is a set of commands and tool adapters (build, git, files, dat), but a command is not a recipe by itself.

What a recipe consists of

• A name. Free-form. Other recipes refer to it.
• Dependencies. Which recipes must run first.
• A body. What the recipe does: takes data, transforms it, puts the result somewhere.

A recipe works with objects, not strings

In Anchor Code, server and client are typed objects. They are not file paths and not a blob of JSON settings. The selected project is supplied automatically, which is why a recipe never says project..

// DRAFT. Shows the idea, does not run.
recipe BuildLcoin {
    dependsOn(GenerateShops);

    GenerateShops.output
        .select(server.shops.lcoin.id)
        .replaceIn(client.dat.shopProducts);
}

How to read it: GenerateShops has to finish first. Then, from its result, the shop that is bound as L-Coin in the selected project is taken and written into the matching client table.

Draft. The unquoted recipe keyword, the dependsOn(...) form, the output property, the select and replaceIn methods and the shopProducts name are not approved. This is a sketch of the meaning, not a contract and not executable code.

What is guaranteed already

• Visibility is not authorization. A recipe that can read data does not thereby get the right to change it.
• Preview comes first, action second. Nothing on the server, in the client, in Git or in production is changed silently.
• If the target is ambiguous, that is an error with an explanation, not a guess.

Next: Your first recipe.