Stretching the strangeness budget to bridge language communities

This series of articles explores differences between programming languages—differences that might make each language the right tool for a different job, but that complicate translating concepts across languages.

What good is a programming language that translates to all the others? If its semantics don't translate faithfully, not a lot. If they do, developers can easily identify common problems and craft shared solutions even across language communities—the software community becomes less divided despite working on different stacks. Data scientists use Python, the web platform uses JavaScript, mobile app and backend people each use their own preferred languages (and legacy systems—ugh). But our different stacks shouldn't divide us.

Temper is a programming language designed for one thing: translating well to the others. It fills this gap in our collective toolbox; it allows sharing common solutions across all the language communities in an engineering organisation, or across the whole open-source community. One developer or a small team can use Temper to write a library that translates into all the other languages, supporting many language communities.

Each article in this series starts with a code fragment and explains problems with straightforward translations of it into various languages. Translating faithfully and providing idiomatic interfaces is hard, so understanding these differences is crucial to bridging language communities and preventing the languages we work in from dictating whom we can collaborate with.

The first article, A Tangle of Strings, starts small. Strings are great for entertaining cats, but did you know they're good for other things too! Can we write simple string functions that translate efficiently while preserving semantics? To do so in a way that is both efficient and sensible to the working programmer, we need to first disentangle some semantic yarn.

Generically Speaking talks about generic operations. Some languages don’t distinguish between strings and integers at runtime (Perl, PHP), others between integers and floats (JavaScript, Lua and many others). How can we translate a simple, generic minimum function so that it performs string comparison given strings and numeric comparison given ints? How can we do this while keeping code size small for languages like JavaScript where code is sent over the network?

When Words Collide deals with naming collisions. Languages have subtly (and grossly) different approaches to naming style, overloading, overriding, masking, keyword reservation. What can one do when the perfect name for a function in one language is forbidden in another?

Well, I Never!, But When I Do discusses bottom types. Some languages have a type that conveys “expressions of this type never produce a value,” which you might've seen called Never or Nothing. How does Temper’s bottom-esque type let us translate well to both languages that have them and those that don’t and avoid producing translations that trip up unreachable-code compiler checks?

Cast Away: TFW Your BFF is a Volleyball discusses type casting, checking whether a value is of a type before proceeding to treat it that way. As noted above, some languages answer “is this value an int, a float, a boolean, or a string” with a resounding “yes!” Other languages represent dates as eight character strings: “20241231” or “tomorrow.” How can Temper provide reliable casting sufficient for important use cases?

Failure Is Not Not an Option?! This article discusses Nicomachean Ethics: Aristotle’s take on the virtues of exceptions vs result checking for error handling. Failure handling is also super important to programmers. An interface isn’t idiomatic if programmers' hard-learned habits don't help them meet their failure-handling responsibilities.

I Love You, You Love Me, We're Surrounded by Uncollected Garbage is about the big purple dinosaur of memory management: reference cycles. Can we give developers used to garbage-collecting runtimes a way to write memory-safe libraries? How does this fit with our “when you do as OCaml does, things just work” design principle? Can developers who aren’t used to thinking about ownership craft interfaces that appear cyclic, like W3 DOM trees?

The Problem With Introspection Is That It Has No End, nor, PKD, does it have portable semantics. But introspection, and its mirror image reflection, are necessary in many language ecosystems for critical operations like serialization. Can Temper support those operations without sinking into a semantic tarpit?

When We Go Low, We Go High talks about an opportunity. Some dynamic languages pair well with a compiled language: CPython with C, JS with Wasm. When translating, can we get the best of both?: idiomatic dynamic language interfaces that delegate most of the work to ahead-of-time optimized code. Can Temper make it easier for an organization gradually adding types to pre-exisitng code in a language that acquired a type system late in its development?