What is the shortest way to represent all php types in string?

Question

I'm writing a universal system that will hopefully one day apply to medicine, etc. (i.e. it's "scientific").

I figure the best way to go about this is to represent all data in php with string (true would be "true", false would be "false", so on and so forth). The reason for this is that there is exactly one string representation of any value in php (e.g. php code itself).

I am posting this question in an attempt to accelerate the design process of this program.

Some values are easily translated to string: numbers, booleans, etc.

Some are not: objects, arrays, resources.

I figured the format for transmitting objects and arrays is basically json, but I'm not sure if this is a tight fit. It's better than what I currently have (nothing), but, at least at some point, I would like to refine this to a point.

Any ideas?

Answer 1

I'm writing a universal system

This is an ambitious goal indeed; so ambitious as to be foolish to attempt.

Now, probably you don't really mean "can do absolutely anything for anyone", but it's relevant to your question that you don't place any limits on what you're trying to represent. That's making your search for a serialization format unnecessarily difficult.

For instance, you mention resources, which PHP uses for things like database connections, open file handles, etc. They are transient pointers to something that exists briefly and then is gone, and serializing them is not only unsupported by PHP, it's close to meaningless.

Instead of trying to cover "everything", you need to think about what types of data you actually need to handle. Maybe you'll mostly be working with classes defined within the system, so you can define whatever format for those you want. Maybe you want to work with arbitrary bags of key-value pairs, in the form of PHP arrays. You might want to leave the way open for future expansion, but that's about flexibility in the format, not having a specific answer right now.

From there, you can look for what properties you want, and shop around:

• JSON is a hugely popular "lowest-common denominator" format. Its main downside is it has no representation of specific custom types, everything has to be composed of lists and key-value pairs (I like to say "JSON has no class").
• XML is less fashionable than it used to be, but very powerful for defining custom languages and types. Its quite verbose, but compresses well - a lot of modern file formats are actually zip archives containing compressed XML files.
• PHP serialization format is really only intended for short-term, in-application purposes, like cache data. Its fairly concise, and closely tied to PHP's type system, but has security problems if users have influence over the data, as noted on the unserialize manual page.
• There are even more concise formats that don't limit themselves to human-readable representations, if that was a relevant factor for you.

Obviously, the list is endless...

Answer 2

I've programmed a solution to this problem. It's a simple class that converts string to int | float | bool | null | string. The idea is that
any value that is not a relativistic value (e.g. an array, something that simply holds other values) is represented by a single string. The implications are broad, I'll do my best to simplify.

Imagine you're making a website, which is basically (and in fact) made of webpages. If a webpage consists of inputs (typically GET and POST form data), and those inputs are string (GET and POST elements are string), all that stands between us and raw php is interpretation of said string.

Or think of it this way: if you model the total potential of php in string, it may not be ultimately how you do things but it works, right here right now. What THAT means is that we can implement it immediately.

The rest of it is left blank, as that is what I mean by "relativistic".

Now, ok, just to cap it all off, if you think about what this implies in form, in the actual php code itself, everything is, at a point at which there is exactly one string per one "non-relativistic" value.

So basically what you have is a bunch of php. The idea is is designed to be semantically AND syntactically as simple and functional as possible (or, at least, completely factorialized). So basically we have one way to represent any potential data in php.

Anyways, you can find it here: https://github.com/cinder-brent/Cinder

Cheers:)

-- edit --

Lo' and behold, I moved the project. It is now at https://github.com/cinder-brent/Leaf