Ikibazo gihuza ibibazo · Hagati Urwego
Nigute Wandika SQL Ibibazo hamwe na AI Ikibazo
Gerageza neza uburyo ushobora guha AI gahunda yawe, gufata gufatanya nabafana na NULL imitego, hanyuma ugenzure neza SQL yakozwe na AI mbere yo kwizera ibisubizo.
Ikibazo 1 Bya 8
Ukurikije umurongo ngenderwaho, niyihe ntambwe imwe iteza imbere ukuri kwa SQL yakozwe na AI?
Reba buri kibazo8 ibibazo
Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.
- Ukurikije umurongo ngenderwaho, niyihe ntambwe imwe iteza imbere ukuri kwa SQL yakozwe na AI?
- Asking the model to be extra careful
- Pasting your schema, such as CREATE TABLE statements and relationships
- Writing your question entirely in uppercase
- Always using the newest model available
- Kuki ugomba kubwira AI imvugo ya SQL ukoresha?
- Only PostgreSQL supports SELECT statements
- AI models refuse to answer without a dialect
- The dialect determines how much disk space a query uses
- Dialects differ in details like date functions and LIMIT versus TOP
- Uhuza amabwiriza yo gutumiza_ibintu hanyuma ukabara SUM (ordre.total). Ni ikihe kibazo gishobora kuba?
- Revenue is inflated because each order total repeats once per item
- The query will always fail with a syntax error
- NULL values are converted to zero
- The results come back sorted in the wrong order
- Kuki umuyobozi atanga inama yo gusaba AI gusobanura ikibazo cyayo mumagambo asobanutse?
- It makes the query run faster
- SQL engines require an explanation before executing
- It exposes misunderstandings, such as counting rows instead of distinct customers
- It reduces the cost of the AI request
- Ni ayahe magambo yerekeye KUBARA (inkingi) na COUNT (*) aribyo?
- They always return the same number
- COUNT(column) skips NULLs, while COUNT(*) counts every row
- COUNT(*) skips NULLs, while COUNT(column) counts every row
- COUNT(column) counts only distinct values
- What is the safest way to experiment with AI-generated queries?
- Run them directly on production during quiet hours
- Ask the AI whether the query is safe and trust the answer
- Use a read-only database user and wrap data changes in a transaction you can roll back
- Remove the WHERE clause so the query is simpler
- When asking an AI to optimize a slow query, what should you provide besides the query?
- A screenshot of the dashboard that uses it
- The database password
- A list of everyone who runs the report
- The execution plan from EXPLAIN or EXPLAIN ANALYZE
- Which of these is a misconception the guide warns against?
- A query that runs without errors must be returning the right answer
- The schema alone is often enough context, without real customer data
- An UPDATE without a WHERE clause affects every row
- Business rules like amounts stored in cents should be stated