symbol-search

An advisory by default: it explains, and blocks nothing, on a text search of the tree for a symbol, a declaration or a diagnostic code the graph answers.

What it catches

A text search of the tree for a symbol, a declaration or a diagnostic code the graph answers.

Why

Each alternative of the pattern is classified by what it looks for, and the deny names every classification and why, so a false positive is disputable from the message alone. A name is a declaration lookup (func X, type X, class X, function X, const X), a CamelCase identifier, or a short name the syntax marks as one: a word search (\bX\b, -w), a qualified member (pkg.X, \.X) or an assignment, when it is capitalized. Text is everything else: a plain word, which is as likely prose; snake_case, which in this tree is a config key, a JSON tag or a Buzz function; a lowercase call or member (Mkdir( is os.Mkdir as often as a local function); a standard-library member (os.Rename); a declaration of a lowercase word many share (func main); a string literal, a path, a hyphenated word or a regular expression. With -i a plain or marked short name widens to text, and a CamelCase one stays a name. The deny fires when ANY alternative is a name the index defines, serving magus refs X --occurrences per name (--definition --source for a declaration lookup, which prints the body in place of the grep-then-sed pair) and magus refs --text <literal> <paths> for each literal text alternative, and magus explain diagnostic:<code> for a code the graph holds a node for. A stale index still answers, serving magus graph build --silent first: it still knows the names it held, and it vouches for a declaration lookup or a CamelCase call it has not indexed yet, since that is the name a branch is adding; a bare word it does not hold stays text. An index goes stale on the first edit, so a rule that stood down for a stale one would let every symbol search through. Only a workspace with no symbol index at all is advised instead, since a deny there routes nowhere. An index built at another revision is stale on the same terms; a rebase still underway turns the deny into graph-stale's advice, since a rebuild then would describe a tree about to move. The search must reach the tree: a directory, a glob, or several files. One named file is a read and runs, with refs advised when the index vouches for the name; a definition lookup carrying -A, -B or -C is a read of the body, and grep-reader refuses it. The files refs answers for are the languages a spell declares a symbol indexer for (here Go and TypeScript), read from the spell catalog. A search of stdin, Markdown, a log, Buzz (source no indexer reads, so a Buzz fun or object lookup is text too), a directory holding none of those languages (a skills tree, fixtures), a dot-directory (.github, .git), node_modules, a revision, or a tree outside the workspace runs. A search of the tree carries the index's own answer when the index is current: every file under the searched paths with its occurrence count and lines, as magus refs prints them. It is refs' answer, not grep's: comments, strings and prose are not in it. A pipe after the search is not reproduced. The deny still carries the unfiltered answer and says so: a model of sort, sed or awk substituted for the real tool diverges from it.

Default and override

By default this rule takes the decision advise. A workspace sets it by name, in its root magusfile, to deny, advise or off:

magus\guard.builtins({"symbol-search": "deny"})

A loosening takes effect once it is committed; a tightening applies at once.

Seeing it

A verdict names its rule in brackets, which is how you got here:

advise [symbol-search]: ...

magus describe rule symbol-search prints the same entry at a terminal, and magus describe rules lists every rule this workspace enforces.

See also

• All rules - what this workspace enforces, deny first
• The guard - how a verdict is reached and wired
• Setting a built-in rule - how a workspace changes this default