DEV Community: Roland HUON

An AI wrote my trading app's currency converter. When it doesn't know a rate, it says 1.0

Roland HUON — Fri, 23 Oct 2026 07:10:32 +0000

On Monday I posted a function Claude Opus 5.5 wrote for my personal trading dashboard and asked whether you'd merge it. Here's the answer.

The function converts every position into my portfolio currency before the dashboard adds them up:

@cached(ttl=3600)
def fx_rate(from_ccy: str, to_ccy: str) -> float:
    if not from_ccy or not to_ccy or from_ccy == to_ccy:
        return 1.0
    # Les cotations londoniennes sont en pence (GBp)
    factor = 1.0
    if from_ccy == "GBp":
        from_ccy, factor = "GBP", 0.01
    if from_ccy == to_ccy:
        return factor
    q = quotes([f"{from_ccy}{to_ccy}=X"])
    rate = next(iter(q.values()), {}).get("price")
    return (rate or 1.0) * factor

It reads well. It knows that London stocks are quoted in pence. It caches rates so I don't hit the API limits. It short-circuits when there's nothing to convert.

It shouldn't be merged.

The problem is a fallback that looks like caution

When the quote provider fails or doesn't know the symbol, quotes() returns an empty dict. Follow it through:

q = {}
rate = next(iter({}.values()), {}).get("price")   # → None
return (None or 1.0) * 1.0                          # → 1.0

No exception. No log. The function answers with a perfectly plausible number.

Line 3 does the same thing in a different place: if a position has no currency, it's treated as if it were already in the target currency.

And then @cached(ttl=3600) makes it worse. If your decorator stores whatever comes back — and most do — that 1.0 stays there for an hour. A quote provider that hiccups for one second gives you sixty minutes of wrong valuations.

What that looks like in a portfolio

Say I hold ¥1,000,000 of a Japanese stock, and my dashboard reports in dollars.

correct:   1,000,000 JPY × 0.0067  ≈  $6,700
fallback:  1,000,000 JPY × 1.0     =  $1,000,000

The position appears about 150 times larger than it is, inside a total that still looks like a real number. On a page whose job is to help me decide what to buy and sell, that's the worst kind of failure: silent, confident, and wrong.

Nobody asked for this, and nobody forbade it

The whole project started from this prompt (my translation; I wrote it in French):

I want to build a platform to organise and plan my trading. I need to know about the news instantly, see what I can invest in, see the confidence index, how many people bought shares in it, the returns, etc. Plan what you need to set up for this.

It says nothing about what to do when data is missing. So the model filled that silence with the most comfortable choice for the code itself: never crash, always return something.

That's the part I keep coming back to. The missing requirement — in a trading tool, a wrong number is worse than no number — is obvious once you say it. I never said it. A better model won't fix that; it will follow my incomplete prompt even more faithfully.

The fix

Fail loudly, and let nothing wrong reach the cache. This is the version I'm moving to:

SUBUNITS = {"GBp": ("GBP", 0.01), "ZAc": ("ZAR", 0.01), "ILA": ("ILS", 0.01)}

@cached(ttl=3600)
def fx_rate(from_ccy: str, to_ccy: str) -> float:
    if not from_ccy or not to_ccy:
        raise ValueError("Missing currency")
    from_ccy, f_from = SUBUNITS.get(from_ccy, (from_ccy, 1.0))
    to_ccy, f_to = SUBUNITS.get(to_ccy, (to_ccy, 1.0))
    if from_ccy == to_ccy:
        return f_from / f_to
    q = quotes([f"{from_ccy}{to_ccy}=X"])
    rate = next(iter(q.values()), {}).get("price")
    if rate is None:
        raise LookupError(f"No quote for {from_ccy}/{to_ccy}")
    return rate * f_from / f_to

Caching decorators generally store return values, not exceptions, so a failed lookup is retried on the next call instead of being frozen for an hour. It's worth checking that yours behaves that way.

The caller changes too: when a rate is unavailable, the dashboard shows "rate unavailable" for that position and marks the total as incomplete. It never adds up a total that's quietly wrong.

A smaller one, with the same ending

Pence only worked in one direction. The original handled GBp as a source currency, but fx_rate("USD", "GBp") returned the USD→GBP rate, 100 times too small. Johannesburg (ZAc) and Tel Aviv (ILA) also quote in sub-units; for those, the symbol most likely doesn't exist, the quote comes back empty, and you're back to 1.0. The fix handles both directions and all three.

What looked wrong and wasn't

"The factor should be 100, not 0.01." It shouldn't. A penny is 0.01 pounds; converting from pence means multiplying by 0.01, then by the pound's rate.

"The second if from_ccy == to_ccy is dead code." It isn't. After GBp becomes GBP, it correctly handles fx_rate("GBp", "GBP"), which should return 0.01.

"Use Decimal instead of float." Reasonable for accounting. For a displayed exchange rate, it doesn't change the outcome.

"A one-hour cache means stale rates." That's what a TTL is for. The problem isn't stale rates, it's invented ones.

What I take from it

Every line here is defensible on its own. The bug is a default: a value the code reaches for when it doesn't know, chosen to keep things running rather than to keep them true.

Worth searching your own code for: or 1.0, or 0, .get(key, 0), an except that returns something plausible. Each one is a question about what your program should do when it doesn't know — and whether anyone ever answered it.

I asked Claude Code to search the rest of my project for the same pattern. That list will probably be next week's article.

I write one of these every week from code an agent wrote for my own projects. If you'd rather catch the next one before the answer: https://discord.com/invite/WsYyZk5sWj

An AI wrote this currency converter for my trading app. Would you merge it?

Roland HUON — Mon, 19 Oct 2026 07:34:50 +0000

I asked Claude Opus 5.5 to build a personal trading dashboard: news, what to invest in, a confidence index, returns. Along the way it wrote the function that converts every position into my portfolio currency before computing totals.

Here it is, unedited:

@cached(ttl=3600)
def fx_rate(from_ccy: str, to_ccy: str) -> float:
    if not from_ccy or not to_ccy or from_ccy == to_ccy:
        return 1.0
    # Les cotations londoniennes sont en pence (GBp)
    factor = 1.0
    if from_ccy == "GBp":
        from_ccy, factor = "GBP", 0.01
    if from_ccy == to_ccy:
        return factor
    q = quotes([f"{from_ccy}{to_ccy}=X"])
    rate = next(iter(q.values()), {}).get("price")
    return (rate or 1.0) * factor

(The comment is in French because I am. It says London quotes are in pence.)

What you need to know

quotes() is the market data helper:

quotes(["EURUSD=X"])   # → {"EURUSD=X": {"price": 1.0874}}
quotes(["XXXUSD=X"])   # → {}  when the provider fails or the symbol is unknown

And this is roughly how it's used (simplified):

for p in positions:
    value = p.qty * p.price
    total += value * fx_rate(p.ccy, base_ccy)

The session started from this prompt, which I wrote in French:

I want to build a platform to organise and plan my trading. I need to know about the news instantly, see what I can invest in, see the confidence index, how many people bought shares in it, the returns, etc. Plan what you need to set up for this.

Your call

You're the reviewer. Merge, request changes, or reject — and leave the one comment you'd actually post on the PR.

Start your comment with your decision in capitals: MERGE, CHANGES or REJECT. If you don't want to spoil it for others, you can hide the reasoning:

CHANGES
{% details my comment %}
your review comment here
{% enddetails %}

One thing to keep in mind: flagging something that isn't actually wrong counts against you. Blocking a good PR has a cost too.

I'll post the answer on Friday, and go through every comment: the right call, the right reason, and the comments that looked convincing but weren't.

It built on Windows, failed on Mac and Linux, and the Dockerfile was fine

Roland HUON — Sat, 03 Oct 2026 13:26:20 +0000

I asked Claude Opus 5 for a Dockerfile and a Compose file for a Spring Boot project. Multi-stage build, dependencies cached in their own layer, a healthcheck on Postgres, the app waiting for the database before starting. I read it, it looked right, I ran it on my Windows machine. It built. It ran. I merged it.

On Mac and Linux, the build died on the first RUN:

/bin/sh: 1: ./mvnw: Permission denied

Here are the lines involved:

COPY .mvn/ .mvn/
COPY mvnw pom.xml ./
RUN ./mvnw dependency:go-offline -B

There is nothing wrong with them. That's the interesting part.

Where the bug actually was

Not in the Dockerfile. In git:

$ git ls-files -s mvnw
100644 bd8896bf2217b46faa0291585e01ac1a3441a958 0   mvnw

100644 means a regular, non-executable file. A script you're meant to run should be 100755. The Maven wrapper had been committed without its executable bit — most likely because the first commit came from Windows, where git doesn't track that bit by default.

On Mac and Linux, checking out the repo gives you an mvnw with mode 644. COPY preserves that mode inside the image. ./mvnw refuses to run.

Why Windows hid it

Windows files don't carry Unix permissions. When you build from a Windows client against a Linux engine, Docker has nothing to copy over, so every file in the build context arrives in the image as rwxr-xr-x. Docker has long printed a warning about exactly this when building from Windows.

So on Windows, mvnw became executable by accident, and the build passed. The one test I ran was structurally incapable of catching the bug.

It wasn't just Docker, either: ./mvnw spring-boot:run failed on Mac and Linux for the same reason. The whole project only worked on Windows.

The red herring

The repo had this .gitattributes:

/mvnw text eol=lf
*.cmd text eol=crlf

It's easy to read that and conclude mvnw is taken care of. It isn't. That rule handles line endings. File mode is a completely separate thing, and nothing in the repo was handling it.

The fix

At the source, so every clone gets it right:

git update-index --chmod=+x mvnw
git commit -m "Make mvnw executable"

And belt and braces in the Dockerfile, so a copy that lost the bit — a downloaded zip, say — still builds:

RUN chmod +x mvnw && ./mvnw dependency:go-offline -B

Or skip the question entirely, since sh doesn't need the executable bit:

RUN sh mvnw dependency:go-offline -B

The same applies to gradlew and any shell script you expect people to run.

Check your own repos

git ls-files -s | grep -E '(mvnw|gradlew|\.sh)$'

Anything in that list showing 100644 is waiting for someone on another OS.

This isn't a rare edge case. Out of curiosity I checked another public Spring Boot project from a different author: same wrapper file, byte for byte, same 100644. Its Docker build was fine, because it used the official Maven image instead of the wrapper. Its local build was not.

What I take from it

The generated Dockerfile was correct. The diff was correct. The review, had there been one, would have looked at those lines and approved them. The bug lived in the relationship between the file and the repository's metadata, which a diff doesn't show — and it only appeared on machines the author never used.

When the code in front of you is right, the questions worth asking are about everything around it: which platform was this tested on, what does the file mode say, what happens on the second machine?

I run a small weekly code-review challenge built around bugs like this one, where every line is correct and something is still broken. This was the first challenge of the season: https://www.linkedin.com/feed/update/urn:li:activity:7510232443304714240. Would you have merged it?

Looking for beta testers: we're building a tool that bundles your trip, hotel and event into one plan

Roland HUON — Wed, 30 Sep 2026 20:29:14 +0000

Hey DEV community 👋

My team and I are a team of students working on Bundle Events as part of an entrepreneurship project at my school, and we're looking for beta testers to help me shape it.

Try it here: https://bundle-event.vercel.app/

The problem

Every time we go to a concert or festival or somehere else in another city, it's the same routine: one tab for train tickets, one for flights, one for car options, another for hotels, and a map to figure out which hotel is actually close to the venue. Then we realize the last train leaves before the show ends. 🙃

What Bundle Events does

You tell it where you're leaving from and where you're going. It then:

compares transport options (train, plane, car)
finds accommodation close to the venue
assembles everything into a single "bundle" you can book
shows the indication of the carbon footprint of each trip, so you can pick the greener option if you want

🇫🇷 France only (for now)

Right now, Bundle Events only covers trips and events in France. That's a deliberate choice: we'd rather make it work really well in one country before expanding.

But if enough of you tell me you'd use it where you live, going international is definitely on the table. So if you're outside France, I still want to hear from you: tell me in the comments which country or city you'd want it in, and which events you usually travel to. That's exactly the kind of signal that will decide where Bundle goes next.

Where we need your help

The project is still early, and we want to iterate based on real feedback rather than my own assumptions. We'd love it if you could:

Create a bundle for a real or imaginary event in France (5 minutes max)
Tell us what confused you, what broke, or what you'd expect to find
Answer a few quick questions here: https://forms.gle/sciKEqaKDpZxtxgQA

Any kind of feedback is welcome: UX, performance, design, the concept itself, or even "I would never use this, here's why". The honest stuff is the most useful.

Tech side

It's built with Next.js, Supabase and deployed on Vercel. If you're curious about how the bundle logic works or want to discuss the architecture, ask away in the comments. Happy to share what I've learned.

Thanks!

Every tester will genuinely help me improve the product (and my project grade 😄). If you know someone who travels a lot for concerts or festivals, feel free to share this with them too.

Drop your feedback in the comments or through the form. I'll try to reply to everyone!