Skip to main content

Your LEFT JOIN Became an INNER JOIN and Nothing Warned You

· 11 min read
Nguyễn Huỳnh Minh Tiến
Middle Fullstack Developer @ Utop.vn
Summary

LEFT JOIN keeps every row from the left table and fills in NULL where nothing matches. But WHERE runs after the join, so any condition placed on a right-hand column also removes those NULL rows — and the LEFT JOIN becomes an INNER JOIN with no warning whatsoever. Measured on PostgreSQL 16: a plain LEFT JOIN gives 4 rows, adding WHERE leaves 2 rows, but moving that same condition into ON gives 3 rows — which is the correct answer.

This is the bug I see most often in reporting queries. It is not a syntax error, it is not slow, it throws nothing. It simply returns too few rows, and usually it is precisely the important ones that go missing: customers with no orders yet, products that have never sold, salespeople who have not closed anything.

This post drills into the JOIN lesson from my 30-day SQL series. Every number below is real output from PostgreSQL 16.11.

Why Does Your NOT IN Return Zero Rows? SQL's Three-Valued Logic and the NULL Trap

· 12 min read
Nguyễn Huỳnh Minh Tiến
Middle Fullstack Developer @ Utop.vn
Summary

NULL in SQL is not a value — it is a marker meaning "unknown". Every comparison against it yields UNKNOWN, and WHERE only keeps rows that evaluate to TRUE, so UNKNOWN is discarded exactly like FALSE. The most serious consequence lands on NOT IN: a single NULL anywhere in the subquery makes the entire query return zero rows, even when the data plainly exists. I ran it on PostgreSQL 16: with the same intent, NOT IN returns 0 rows while NOT EXISTS returns 3. No error, no warning — the query just quietly returns the wrong answer.

This is not the kind of bug that crashes an application. It just makes a report short by a few numbers, makes a list screen empty, makes a campaign miss some customers — and leaves nothing behind in the logs.

This post drills into one detail from my 30-day SQL series, specifically the lessons on operators and expressions and subqueries. Every number below is real output from PostgreSQL 16.11.

One Vietnamese Diacritic Killed an API Call: cf-ipcity, HttpClient and the ASCII Limit

· 11 min read
Nguyễn Huỳnh Minh Tiến
Middle Fullstack Developer @ Utop.vn
Summary

Cloudflare injects cf-ipcity into incoming requests, and for visitors in Vietnam the value is Hồ Chí Minh — with diacritics, which means non-ASCII. Our .NET service forwarded every incoming header verbatim onto its outgoing calls, so that value landed in HttpClient. The surprise is that Headers.Add does not throw, and TryAddWithoutValidation returns true; everything only blows up at SendAsync with HttpRequestException: Request headers must contain only ASCII characters, and not a single byte leaves the process. The bug is neither Cloudflare's nor .NET's — it is in an application that forwards every header unconditionally.

The setting is an e-commerce loyalty platform serving roughly three million customers. I have removed the client's name and every identifying detail; what remains is the technical part.

Everything looked normal. The API was running. The Kubernetes pods were healthy. The database was fine. Requests were reaching the application. But one HTTP call to an internal service kept failing in production — and only in production.

What broke it, it turned out, was the name of the user's own city.

Forwarding Headers in ASP.NET Core: Why 'Forward Everything' Is an Architectural Bug

· 12 min read
Nguyễn Huỳnh Minh Tiến
Middle Fullstack Developer @ Utop.vn
Summary

Copying every incoming header onto an outgoing request is an anti-pattern, and it fails in three different directions: correctness (one non-ASCII value makes HttpClient throw), security (you trust and relay data the client set itself), and architecture (headers injected by infrastructure become part of your application's contract). The fix is an allowlist — an explicit list of headers permitted to pass — placed in a shared DelegatingHandler. When you write that handler, mind one trap: it does not live in the request's scope.

This post closes a three-part series that began with a production incident on an e-commerce loyalty platform of roughly three million customers, where Cloudflare's cf-ipcity: Hồ Chí Minh header broke an internal call. The first part told the story, the second explained why headers cannot carry Vietnamese. This one answers what is left: how do you rewrite that code correctly?

Can an HTTP Header Carry Vietnamese? ASCII, obs-text, and Where .NET Draws the Line

· 11 min read
Nguyễn Huỳnh Minh Tiến
Middle Fullstack Developer @ Utop.vn
Summary

Short answer: no, not if you want it to survive every layer. The HTTP specification defines header values in terms of printable ASCII, plus an obsolete branch called obs-text that allows bytes 0x80–0xFF but defines no charset for interpreting them. The consequence is that a Vietnamese string can get through partially: í lives inside Latin-1 and is representable, while ồ is not, because it is U+1ED3 — beyond 0xFF. .NET takes a decisive stance: HttpClient refuses to send any non-ASCII header value. To carry text with diacritics you must encode it, using percent-encoding, RFC 8187 or Base64.

This post branches off a production incident where the header cf-ipcity: Hồ Chí Minh, injected by Cloudflare, made HttpClient throw. There I stopped at the symptom. Here I answer the question that symptom raises: what exactly is an HTTP header allowed to contain?