Thread: Read + Write = ?

I honestly can't see the problem with "Serialize", or "RW" or "IO".
Why the OP is choosing "Literate" can only mean one of two things:

1. They have no idea what they are doing and didn't get the joke
2. It just so happens they are serializing the alphabet or dictionary terms. Which is still not a good idea.

In any case, referring back to post #7 (and assuming I understood the requirements):

ObjectLiterator (for both ObjectWriter and ObjectReader),
LiteratorDispatcher (ReaderDispatcher and WriterDispatcher), Literator (Reader and Writer) and LiteratingMapper (WritingMapper and ReadingMapper) seems just right.

Are you really, really, really sure you want to do this?
Odds are you will want to eventually break the two actions apart. It will probably become inevitable as soon as you start adding to your serialization framework. Instead of simplifying ahead of time, I suggest you perhaps leave read and write operations separated. 1) Because they are indeed different actions even if they share code, 2) Because almost certainly they will stop sharing code sometime in the near future and 3) because users of your framework expect different calls for different actions.

There's also such a thing as premature simplification. I suggest you do not fall for it.

Unfortunately, they are easy to be separated into two methods or even two classes.

Btw I didn't get the joke since in my native language dictionary "literate" means both reading and writing activity.

Forgive me for my stupidness.

Uhm..

Hehe, in about any language it should mean read and write. That's precisely the beauty of the joke.
But for matters of human literacy, not for computer reading and saving of data.

Originally Posted by Mario F.

But for matters of human literacy, not for computer reading and saving of data.

Ixnay. You are just making up that first clause, as if there were traditionally some other possible use for the word -- ie, it certainly doesn't say in my (or ANY) dictionary: this word literate applies only to human endeavour.* Audinue's use of it is just hunky dory IMO.

But I think I see Mario's joke. The word "literate" (lit-er-it) is a noun (or adjective) in English and refers to someone who can read or write. However, we also have a verb, transliterate, which means "to write a word in the characters of another alphabet". Serialization is not that, that would be more a data2txt activity.

You could drop the "trans", tho, and use "literate" (lit-er-ate) as a verb to mean "copy". It's also almost "iterate" which could mean, in this sense with the "l" , to read back.

So I totally like the word here.

* it may say "person" but if you taught an ape to read you would call it a literate ape. So really it applies to beings or just things, in so far are that makes sense, which it does WRT a "lit-er-it" computer process that "lit-er-ates".

MK27, your mention of the word "transliterate" made me think of a term that might be appropriate: transduce. It means "to convert from one form to another form," which is perfectly apt for both serialization and deserialization.

Originally Posted by brewbuck

MK27, your mention of the word "transliterate" made me think of a term that might be appropriate: transduce. It means "to convert from one form to another form," which is perfectly apt for both serialization and deserialization.

But serialization does not involve any form of conversion, which is why I said the "trans" prefix in inappropriate. Serialization is a literal, one to one, exact duplication of data.

Originally Posted by MK27

But serialization does not involve any form of conversion, ...

Serialization has two abstraction layers at least.

1. Structure Definition
2. Data Mapping

The first layer defines the object structure.

The second layer maps/converts memory representation of data into a specific format, for instance Binary format, XML format and URL format (they are supported in my library ).

Serialization is yet another data conversion I think and yes, it involves conversion.

Originally Posted by audinue

Serialization has two abstraction layers at least.

Okay. I did not realize the term "serialization" could be applied to, eg, saving data in an XML (text) format.*

I was under the impression that serialize meant to write a segment of memory, such as a struct, directly to a ("binary") file, so there is no kind of translation or conversion. For example, if you have a 144 byte struct, writing 3 of them to a file would create a 432 byte file. This does not involve any transformation.

*handy function tho

When speaking of the serialization process in general it never occurred to me (and I can't recall anyone else doing it either) to specify both the tasks of serializing and deserializing, unless there is a real necessity to do so. The latter is just implicit in the more general term we known as Serialization and that we use in common language.

So, were I to have a need to merge both read and write operations for the purposes of serialization and deserialization, I wouldn't think twice calling it Serialize or some form of it. Searching for some other word seems to me confusing, not to the point and a unnecessary exercise. IO and or ReadWrite seems convenient too, especially if the merged functions referred to the strict actions of reading and writting data to a medium. But Transduce? Literate?... Exchange? Really, enough joking.

As for mapping, MK27 is correct in my option. There's not requirement for serialization to perform any sort of data transformations. For all purposes, one can serialize an object by simply writing its current memory representation byte by byte. If there's any data mapping occurring, it will be a decision of the programmer usually to store it in a more convenient format for quick retrieval later or to be accessible by other programs. But definitely not anything deemed necessary to perform the task of serializing an object.

Originally Posted by Mario F.

As for mapping, MK27 is correct in my option. There's not requirement for serialization to perform any sort of data transformations. For all purposes, one can serialize an object by simply writing its current memory representation byte by byte. If there's any data mapping occurring, it will be a decision of the programmer usually to store it in a more convenient format for quick retrieval later or to be accessible by other programs. But definitely not anything deemed necessary to perform the task of serializing an object.

Tell me how do you map trees and graphs byte by byte? Especially the circular one.

Originally Posted by audinue

Tell me how do you map trees and graphs byte by byte? Especially the circular one.

Tell me how you may an int.

Do you have a problem reading, or don't you really understand the meaning of the words "It's not a requirement"?