SQL editor and completion

Query tabs

• New query tab: Cmd+T (or Cmd+N, File > New Query). It is bound to the connection in use; with none, the tab opens unbound and the connection chooser opens.
• Each tab keeps its own connection, shown on the toolbar. Switching tabs switches the target; nothing you click in the Explorer redirects a query.
• Split the editor: Cmd+\ (right) or Shift+Cmd+\ (down). Each group has its own tabs.
• Switch Tab (Ctrl+Tab) lists the query tabs with their connection and environment, plus the result tabs.
• Results belong to the tab that ran them: switching tabs shows that tab's last run.

SQL files

File > Open... (Cmd+O) opens a file by what it is: a .sql, .txt or .md file as a tab bound to the file, a .lrpt as a report, a SQLite database file as a connection. Text files up to 8 MB open; binary files are refused. File tabs save with Cmd+S. A dot marks unsaved edits. When another program (or a Git checkout) changes the file, a clean tab reloads; a tab with edits offers Compare, Keep mine or Reload and is never overwritten.

Query tabs that are not files are kept by your workspace.

Completion

Completion pops up as you type identifiers or ., and on demand with Ctrl+Space. Enter or Tab accepts, Esc closes, Up / Down move. Accepting is one undo step.

What it proposes depends on where the caret is:

After	Proposed
alias. or table.	the columns of that table only
FROM, JOIN, INTO, UPDATE, TABLE, a comma in a FROM list	tables and views
SELECT, WHERE, ON, SET, BY, AND, OR, HAVING, (	the columns of the tables already in the statement, with nothing typed
anywhere else	columns of the statement's tables first, then aliases, tables, functions and keywords
inside a string or a comment	nothing

• Aliases are read from the statement: FROM t a, FROM t AS a, JOIN t a, UPDATE t, INSERT INTO t.
• Keywords follow the case you type: sel gives select, SEL gives SELECT. Functions are inserted with their opening parenthesis.
• Ranking: the best text match always comes first (prefix, then word start, then substring). Among equal matches, what you use more often ranks higher, and after FROM orders JOIN the tables you have joined with orders before come first.
• What it learns stays on your computer, per connection: object names and counts only, never values or results. It learns from statements you write, statements you run and proposals you accept.
• Completion never queries the database while you type: it uses the structure Lume has already read. View > Refresh Explorer (Cmd+R) reads it again after you change the schema.

JOIN conditions

Right after JOIN <table> ON, Lume proposes the whole condition:

SELECT *
FROM shipments s
JOIN order_items oi ON oi.order_id = s.order_id AND oi.line = s.item_line
--                    ^ proposed from the foreign key, both columns of the composite key

1. Foreign keys first, between the joined table and each table before it, in either direction, with the aliases of your statement and every pair of a composite key.
2. Then same-named columns that are a primary key on one side (p.customer_id = o.customer_id), labelled "same column name". Generic columns such as created_at, and each table's own id, are never proposed.

Foreign keys are read once per table, the first time a JOIN needs them.

Ask the assistant

With Lume Intelligence set up, Cmd+K in a query tab opens the assistant about that tab: describe a change in words and it comes back as a proposal you apply or reject. See SQL assistant.