Add conversion from Reader to InputStream

[Marcono1234]

It appears there are use cases where one wants to convert a java.io.Reader to a java.io.InputStream using a given charset, see this Stack Overflow question or this Baeldung article.

Guava already has this functionality implemented as ReaderInputStream, but only exposes it for CharSource (therefore not for Reader), see also #642 (comment).

Would it make sense to expose it also for Reader, e.g. as com.google.common.io.CharStreams.asInputStream(Reader, Charset)?

Considerations:

• Users might want to specify buffer size?
• Users very likely want to specify behavior on unencodable chars, currently ReaderInputStream uses CodingErrorAction.REPLACE for them

[pt657407064]

  public static InputStream asInputStream(Reader reader, Charset charset, int bufferSize,
                                          CodingErrorAction newAction) {
    return new ReaderInputStream(reader,
            charset.newEncoder()
                    .onMalformedInput(newAction)
                     .onUnmappableCharacter(newAction),
            bufferSize);
  }

Something looks like this maybe? @Marcono1234

[Marcono1234]

Yes that looks reasonable, but it would probably be good to have an asInputStream(Reader, Charset) method for simple use cases as well.
Though I am not sure if this feature is actually something the maintainers want to add.

[Marcono1234]

@ManishOffi, I am not a member of this project, so I cannot assign you, sorry.

[louiscb]

Hi @Marcono1234 & @pt657407064 is this issue still relevant? I'd like to contribute and try to fix it with a PR.

[eamonnmcmanus]

It is possible for anyone to do this with the existing API, like this:

public static InputStream asInputStream(Reader reader, Charset charset) throws IOException {
  return new CharSource() {
    @Override public Reader openStream() {
      return reader;
    }
  }.asByteSource(charset).openStream();
}

Not very elegant, but the com.google.common.io API does put the emphasis on reusable sources and sinks rather than one-shot streams like Reader and InputStream.

[louiscb]

Great thanks @eamonnmcmanus. Can I be assigned to the issue?

[eamonnmcmanus]

The Guava team has talked this over and we don't think there's enough evidence of demand for such a method. The com.google.common.io package has existed at Google for 15 years and we seem to have got by just fine all that time without the method. People who do need it for some reason can do what I suggested above.

[kevinb9n]

I think also that the functionality is slightly unusual. Normally encoding is part of writing, and decoding associated with reading. (Where "encode/decode" really mean "transcode from/to utf-16".) It's enough that I would at least inquire into how the user got into that situation.

[bjmi]

@eamonnmcmanus Unfortunately CharSource.AsByteSource.openStream() creates a ReaderInputStream that silently replaces and thus ignores conversion errors. This could lead to corrupt data, especially bad if it's not detected over a longer period of time.

guava/guava/src/com/google/common/io/ReaderInputStream.java

Lines 87 to 88 in 0a17f4a

	.onMalformedInput(CodingErrorAction.REPLACE)
	.onUnmappableCharacter(CodingErrorAction.REPLACE),

It would be great if there is an overload of CharSource.asByteSource(...) with CharsetEncoder. Alternatively ReaderInputStream could be made public then a CharsetEncoder can be passed.

guava/guava/src/com/google/common/io/ReaderInputStream.java

Line 101 in 0a17f4a

ReaderInputStream(Reader reader, CharsetEncoder encoder, int bufferSize) {

@kevinb9n [...] It's enough that I would at least inquire into how the user got into that situation.

ReaderInputStream from Apache Commons explains it quite good. One uses a third-party API that requires an InputStream and the actual data is character based like a String or a Reader and just an adapter is needed.

Note that while there are use cases where there is no alternative to using this class, very often the need to use this class is an indication of a flaw in the design of the code. This class is typically used in situations where an existing API only accepts an InputStream, but where the most natural way to produce the data is as a character stream, i.e. by providing a Reader instance. [...]

Guava should provide such an adapter too.