Some classes that perform Input/Output (LengthOf, BytesOf, ...) close the stream

[fabriciofx]

Some classes as LengthOf closes the Input after performed the operation. But often we need keep the Input stream opened even after it. So, let's create a decorator to Input and Output that allows keep these streams opened.

[victornoel]

@fabriciofx or we could also just not close those Input or Output in any class and introduce some concept to wrap operations that manipulates Input/Output to close them at the end (a bit like try-with-resource).

I'm not sure what is best…

edit: see #1449 (comment)

[fabriciofx]

@victornoel I think not close after the manipulation better. As say the golden rule: *"Who allocated the resource is the responsible to deallocated it". A try-with-resourced kind of class is a good idea too. :)

[victornoel]

@fabriciofx actually, after some thought, I'm not sure, because when you call Input.stream(), it means you could retrieve a new stream, so you HAVE to close it at the end.

Then, it means your first proposal was actually best indeed: i.e., it is the Input object creator that decides how the streams returned from stream will behave.

[victornoel]

@0crat in

[victornoel]

@fabriciofx some extra requirements for this job for the DEV that will take it: we must find all the classes that closes the stream like this, document them about this behaviour and refer to the existence of the new classes we should introduce that prevent stream closing

[baudoliver7]

@victornoel Let's me find first all the classes that close the stream like mentioned by @fabriciofx, document them properly and add a todo. Isn't it ?

[victornoel]

@baudoliver7 you can do it either order you want yes: first document then todo for implementing the new decorators, or first implement the new decorators and then todo to document the classes.

[baudoliver7]

@victornoel Ok. Let's us implement the new decorators :)

[baudoliver7]

@victornoel What names could you suggest me ?

[victornoel]

@baudoliver7 hehe, no idea actually, @fabriciofx would you have an idea?

Could you propose a name and we discuss it in the PR maybe?

[baudoliver7]

@victornoel Yes, I will do it.

[andreoss]

@fabriciofx
Keeping OutputStream opened might be handy, but what is the point of doing so for InputStream?

First invocation of new LenghtOf(new SafeInput(input)) will return correct length, the second will return 0 in your case, instead of throwing an exception (current behaviour). SafeInput should have some mark/reset logic at close for the original stream in order to avoid that.

Also LengthOf sometimes is explicitly used as a terminal operation on streams (with TeeInput for example), so if LengthOf won't close its streams existing code will break

[baudoliver7]

@andreoss
You are right. The first invocation of new LenghtOf(new SafeInput(input)) will return correct length, the second will return 0 in our case. But we fixed it in the last commit. Thanks.

[baudoliver7]

@andreoss Also, if someone want LengthOf to close original streams, he should not have to decorate original Input with SafeInput. But in this case, the class that allocated the resource is responsible to deallocated it.

[victornoel]

@andreoss @baudoliver7 it is normal that using LengthOf twice on a SafeInput will have the second time return 0, we don't want to prevent this. And actually, you don't know what the underlying stream is made of: maybe you read it once, then later on, there is again some data in it. That's why you need something like SafeInput, to prevent the stream consumer to close it when, as the user (or maybe creator) of that stream, you do know its behaviour is a bit more tricky than that.

This also applies to ouput: when I am producing a Zip with ZipOutputStream, I dont want to close the stream inbetween writing each file for example.

The whole point is to give more power to the code that composes the various objects together, because that code may know better than the code consuming the stream.

[0pdd]

@fabriciofx the puzzle #1577 is still not solved.

[victornoel]

@fabriciofx sorry, this was automatically closed because the "closed" keyword was in the commit message ^^ I hope the solution is good for you, if not please reopen a ticket :)

[0crat]

@rultor/z the issue is closed not by @fabriciofx/z (its creator); I won't close the order; please, re-open it and ask @fabriciofx/z to close it

[victornoel]

@0crat out

[0crat]

@sereshqua/z please review this job completed by @baudoliver7/z, as in §30; the job will be fully closed and all payments will be made when the quality review is completed

[0crat]

@0crat out (here)

@victornoel Job gh:yegor256/cactoos#1449 is not assigned, can't get performer

[sereshqua]

@0crat quality good

[baudoliver7]

@0crat status

[0crat]

@0crat status (here)

@baudoliver7 This is what I know about this job in C63314D6Z, as in §32:

• The job #1449 is not in scope
• Job footprint (restricted area)

[0pdd]

@fabriciofx the puzzle #unknown is still not solved.