query --name-like
Case-insensitive substring search over symbol names.
Read this when you do not know the identifier yet. It does a case-insensitive substring search over symbol names, so a fragment such as SVG finds the types that matter.
How to run it
Full flag list: query in the CLI reference.
Flags
| Flag | Argument | Meaning | Default |
|---|---|---|---|
--name | <symbol-name> | ||
--db | |||
--json | |||
--case-sensitive | |||
--exact |
Needs: index
Pitfalls
- a project DB IS the compile closure, so a hit is membership and a MISS is non-membership.
- STRING LITERALS, DFM, SQL and COMMENT PROSE -- // { } (* *) and ///. Filter with --kind: literal|const|resourcestring|format|comment|doc|dfm-prop|sql-exception. --kind literal is what this returned before comments were indexed.
Details
Case-insensitive substring search over symbol names. This is the discovery query -- the one you reach for when you do not know the identifier yet.
Every other drag-lint lookup assumes you already have the name. This one gets you to it.
The question it was built for
Does this Delphi install have anything that can rasterise an SVG?
query --name SVG answers that with two rows, both literally named SVG.
The types that actually mattered -- Vcl.Skia.TSkSvgBrush, FMX.Skia.TSkSvg,
dxSVGCore.TdxSVGBrush -- were previously found by guessing candidate names one
at a time, or by grepping the RTL and DevExpress trees.
drag-lint query --name-like svgbrush --db C:\Projects\.drag-lint\library-Win32.sqlite --limit 5kind name qualified_name
---------------------------------------------------------------------------
field FSvgBrush Vcl.Skia.TSkSvgGraphic.FSvgBrush : TSkSvgBrush
class TSkSvgBrush FMX.Skia.TSkSvgBrush
class TSkSvgBrush Vcl.Skia.TSkSvgBrush
class TdxSVGBrush dxSVGCore.TdxSVGBrush
5 match(es)Running it from the CLI
drag-lint query --name-like <substring> [--kind class,interface,...] [--limit N] [--json] --db <file.sqlite>Reaching it in the IDE
No IDE surface -- this is a CLI-only feature.
What it needs
An index IS required. Resolve it with
drag-lint resolve-dbs --project <X.dproj> or --in <X.pas>. For API discovery
the library index is usually the one you want, since that is where the RTL,
VCL, FMX and third-party symbols live.
How it differs from --name
| matches | |
|---|---|
--name Foo | the symbol named Foo, plus an edit-distance fallback for near-misses |
--name-like Foo | every symbol with foo anywhere in its name |
They are separate flags on purpose. Quietly widening --name into a substring
search would change what every existing caller gets back, the IDE plugin
included.
The edit-distance fallback is not a substring search and cannot stand in for
one: SVG does not fuzzy-match TSkSvgBrush, because the candidate is four
times longer than the query. Nor does query --text, which searches string
literals -- captions, constants, SQL messages -- and cannot see a type name at
all.
Ordering and the cap
Results come back shortest name first, then alphabetically. For a discovery
question TSkSvg is a far more useful first row than
TdxSVGImageCollectionHelperInternal.
--limit defaults to 50. When it truncates it says so:
-- LIMIT REACHED at 50; there may be more. Pass --limit N, or narrow with --kind.--kind takes a comma-separated list (`class,interface,record,type,function,
procedure,field,...`) and is usually the better way to cut noise, because
discovery is nearly always "what class does X".
Speed, and the one case that is slow
The search is driven by the symbol_trigrams table rather than a bare
LIKE '%x%', which cannot use an index because of the leading wildcard.
Measured on library-Win32.sqlite (3.3 GB): 18,923 ms for the bare scan
versus 12 ms driven from trigrams.
A term shorter than three characters has no trigram, so it falls back to that full scan. The command tells you before the wait:
NOTE: "cx" is shorter than 3 characters, so the trigram index cannot narrow it
-- this falls back to a full scan and will be slow on a large index.JSON
--json emits the same row shape as query --name, with one difference worth
knowing: rows are labelled "match_kind": "substring". They are not
"exact" -- exact is a claim that the symbol is what you asked for, and
qlitesymbol is not the name of TSQLiteSymbolStore.