发送文件更好或输出流

Question

In my project we are saving some files. When a file is requested, I fetch the file from the HBase table and then send it to the browser.

Data is saved in Bytes in HBase. So when I fetch the file I am getting it in Bytes.

Now I have to convert these Bytes into either File object or in Output Stream.

What I want to know is whether send the file via a File object is better or Output Stream?

The file can vary from 1 MB to 10 MS's.

I am using REST API to send data.

Answer 1

It is not entirely clear what you are doing, but the truth is that a java.io.File is just an abstraction over a path, and therefore it contains no data whatsoever. As such, I do not think that sending a java.io.File object would help you in any way if what you want to do is sending the actual file contents.

Neither it makes sense to send an OutputStream which is only a class to control the flow of bytes into a given sink. What you send are the bytes, not the stream (ie if your sink were a socket, then it makes sense to send the bytes of the contents of the file through the socket using the OutputStream associated with this given sink).

Answer 2

Sending a java.io.File object wont help as it would be a very java specific binary object and it will mandate that receiving system is in java.

You mentioned, you are using REST . Now this calls for something which is independent of any technology/platform .

One feasible approach would be as follows.

1. Let the rest API return a URL which points to the file user wish to download.
2. Write/Configure a Servlet for this URL.
3. When this servlet receives the request, it reads the database and writes to the response stream (output stream).
4. The user will get a prompt asking for file download.

Here is an example of PDF file download using a servlet .

Answer 3

I'm assuming that the question is asking whether to write directly to an OuputStream or to save the byte[] to a local file first. The only reason I could think of to save to a local file first would be to decouple the retrieval from HBase and the user's download. If you expect each request that returns a byte[] from HBase to return the whole file to the user in the response, use OutputStream.

Having said that, retrieving 10 MB as a byte[] seems like it might consume a lot of memory per request. You might want to see if you can get HBase to give you an InputStream so you don't need up to 10 MB RAM per request.