Reading a Query Before You Run It

#Read It Like a Sentence

When you get a query from someone else, do not start by running it. Read it first, in the order a database would act it out. Ask four small questions, one clause at a time.

1. FROM: which table does it start with?
2. WHERE: which rows does it keep? What exactly is the test?
3. SELECT: which columns will I see?
4. ORDER BY and LIMIT: in what order, and how many rows?

Try it on this query.

SELECT name, hire_date
FROM employees
WHERE hire_date >= '2021-01-01'
ORDER BY hire_date DESC
LIMIT 3;

Do: Compare the three names you predicted with the result. Then change LIMIT 3 to LIMIT 4.

Notice: Jack Taylor, Grace Lee and Eve Brown come first. With LIMIT 4, David Wilson (2021-02-10) appears as the fourth row.

Why: The test hire_date >= '2021-01-01' keeps four rows. LIMIT 3 was cutting the last one, and a reading would have told you that before you ran anything.

In one sentence: read each clause in order, say the query as one plain sentence, and predict a few rows by hand before you run it.

#The Missing Value Trap

• Bob: 1 is not different from 1, so the test is false. Dropped.
• Eve: 3 is different from 1, so it is true. Kept.
• Grace and Jack: 6 is different from 1, true. Kept.
• Iris: 3, true. Kept.
• David and Henry: 1, false. Dropped.
• Alice, Carol and Frank: the manager is NULL. Is "no value" different from 1? The database cannot say yes or no, because it has no value to compare. It treats the answer as unknown, and WHERE keeps a row only when the test is true. Unknown is not true, so these three are dropped too.

The result is four rows: Eve, Grace, Iris and Jack. Many people expect seven, because the three people with no manager surely do not report to manager 1. The database does not guess.

Do: Run it and count the rows. Then change the last line to WHERE manager_id = NULL; and run again. Then change it to WHERE manager_id IS NULL;.

Notice: The first run gives 4 rows: Eve, Grace, Iris and Jack. The = NULL run gives 0 rows with no error. The IS NULL run gives 3 rows: Alice, Carol and Frank.

Why: A missing value is never equal to anything, not even to another missing value, so = NULL is never true. The special phrase IS NULL is the test that can find missing values.

To get the seven people you probably wanted, say both parts out loud: "not under manager 1, or with no manager".

SELECT name, manager_id
FROM employees
WHERE manager_id != 1 OR manager_id IS NULL;

That returns seven rows: Alice, Carol, Eve, Frank, Grace, Iris and Jack.

In one sentence: NULL means no value, an equals or not-equals test on it is never true, and IS NULL is the only test that finds it.

#Check the Count

A result can look tidy and be wrong. Two cheap checks catch many problems. Neither needs any new SQL, only a little arithmetic you can do with the table in view.

Do: Run the query as it is. Then change 'engineering' to 'Engineering' and run again.

Notice: The first run gives 0 rows and no error message. The second gives 4 rows: Alice, Bob, David and Henry.

Why: A different capital letter is a different word to the database. No error only means the database could follow your words, not that they matched anything.

In one sentence: guess the count first, treat zero rows as a clue, and test one row you can check by eye.

#Checking SQL That an AI Tool Wrote

Many people now ask a chatbot to write SQL from a plain question. This is a real time-saver, and it is also where this lesson pays off. An AI tool writes a query that looks right. It does not run your database or know your data. Read its work the way you read a colleague's: carefully, and before you trust it.

Here is a plain question: "Who are the top 3 paid people in Engineering?" Suppose a tool answers with this.

SELECT name, salary
FROM employees
WHERE department = 'Engineering'
ORDER BY salary
LIMIT 3;

Act it out. The four engineers by salary are Bob 85000, David 92000, Alice 95000 and Henry 98000. Smallest first and keep three gives Bob, David and Alice. Henry, the top earner, is missing, and the answer looks completely reasonable.

Do: Run it. Then add DESC after salary in the ORDER BY line and run again.

Notice: Without DESC you get Bob 85000, David 92000 and Alice 95000. With DESC you get Henry 98000, Alice 95000 and David 92000.

Why: ORDER BY sorts smallest first unless told otherwise, and LIMIT keeps the start of the sorted list. One missing word flipped the question the query answers.

Use this short checklist on any query you did not write.

1. Does it only ask? Check for words that change data (the next section).
2. Right table and columns? Compare the names with the real table.
3. Does the filter say what the question says? Check each test: greater or at least, text spelled and capitalised as stored, and what happens to empty cells.
4. Do the sort and limit match the words "top", "latest", "first"? Top usually needs DESC.
5. Does one hand check agree? Pick a row you can verify and see whether the result agrees.

None of this needs you to write the query yourself. It needs you to read it, which is a smaller skill and one you now have.

In one sentence: treat an AI-written query as a draft, read it with the checklist, and check one result by hand.

#Words That Change Data

Every query so far only asked. Some queries do not ask. They change what is stored, and you cannot take that back by pressing Run again. A few words give them away.

If a query contains one of these words, stop and read twice. Here is what a missing filter can do:

UPDATE employees SET salary = 0;

In one sentence: INSERT, UPDATE, DELETE, DROP and ALTER change data, so a query with one of them is read twice and tried on a copy before it touches real work.

#Where to Go Next

You can now read a query, spot the missing-value trap, and check a result before you trust it.

• Next: Your First SQL Query walks through SELECT, columns and LIMIT in more detail, with the other sample tables.
• Then: Filtering & Sorting Data covers AND, OR, NOT and sorting properly.
• Later: INSERT, UPDATE, DELETE: Modifying Data teaches the words from the last section, with transactions as a safety net.

In one sentence: carry the reading routine and the checklist into every lesson that follows.

#Key Takeaways

• Read a query clause by clause, in the order a database acts it out, and say it as one plain sentence before you run it
• NULL means no value. An equals or not-equals test on it is never true, so rows quietly drop out, and IS NULL is the test that finds them
• Guess the row count first, treat zero rows as a clue, and check one row you can verify by eye
• An AI-written query is a draft. Check the filter, the sort direction and the limit against the words of the question
• INSERT, UPDATE, DELETE, DROP and ALTER change data, so read those twice and try them on a copy

#Quick Check