commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hariharasudhan R" <hari.develo...@gmail.com>
Subject Re: [compress] Interface is ready
Date Wed, 05 Apr 2006 16:27:43 GMT
Hi Sandy,

Does this work with Solaris tar, since Solaris tar is different from GNU tar
?
If not , is there a plan to do so ?
Won't making TarEntry an interface with Solaris/GNU implementations be a
good idea ?
This is just an idea off the top of my head.. I've not delved too much into
the code..

Thanks,
Hari.

On 4/4/06, C. Grobmeier <grobmeier@possessed.de> wrote:
>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hello Sandy,
>
> thousand thanks for your comments and the time you have spent for this.
> I will work through this all and prepare another draft soon.
>
> Thank you very much,
> Chris
>
> > The following is a stream of things I don't think are right in the
> > order I observe them while skimming the code. I've never used the
> > compress code so excuse any suggestions that miss the point.
> >
> > * Archiver: too many overloaded methods. Especially with setters.
> > Having setFoo(File) and setFoo(String) will breaks the JavaBean
> > property conventions and could cause problems with scripting tools. If
> > you're going to do something like this then use different names eg:
> > setDestinationFile(File) and setDestinationFilename(String) and let
> > the latter be a virtual property that really just creates a new File
> > and calls setDestinationFile(File).
> >
> > * Archive: doesn't do much, only contains one static method:
> > getInstnace(ArchiverType). Because it's static you cannot subclass it
> > so there is no way to extend it's behavior for when someone else want
> > to implement CabArchiver. The type-safe enum example I gave in the
> > other email makes this class unnecessary.
> >
> > * ArchiveType/CompressorType: make them type safe enums. Actually I
> > think it's a mistake to distinguish between Archive types and
> > Compression types. Making them separate you have one library that does
> > two things and I think you should treat them as one thing and do that
> > one thing well or have two libs that each do one thing.
> >
> >
> > * TarConstants: is package scoped and, with the exception of one line
> > in TarOutputStream, is only used by TarEntry. It looks to me the logic
> > in TarOutputStream that uses TarConstansts could be easily pushed into
> > TarEnrty and then the constants in TarConstants should be moved to
> > where they are used.
> >
> > * TarInputStream: skimming it I don't it. It looks like it's taking
> > the InputStream concept and corrupting it with the notion of many
> > separate streams for each file in one stream. This is confusing
> > because it doesn't fit the expectations of an InputStream. IMO it
> > should be it's something similar to an Iterator that takes a raw
> > InputStream and provides a way to get at metadata sequentially from
> > the raw InputStream. From the meta data you should be able to get an
> > InputStream which is just for that file in the archive.
> >
> > * TarOutputStream: Same problem as TarInputStream but with an
> > OutputStream. Because TarOutputStream subclasses OutputStream it makes
> > write methods available. But using these methods are dangerous because
> > if you write a different number of bytes than what was passed when
> > putNextEntry(TarEntry) was called you corrupt the archive. A good API
> > doesn't let the programmer make mistakes. I'd change this to be it's
> > own object type that accepts TarEntrys with all the needed to add a
> > file in one step that either succeeds or fails.
> >
> > * TarInputStream/TarOutputStream: I don't see why these classes should
> > be public. Making them package private would make me care less what
> > they are because they aren't part of the public API and can be changed
> > at will.
> >
> >
> > * ZipLong, ZipShort: these are both public, and I don't see why they
> > need to be. They are only used inside the package with the exception
> > of 3 places ZipShort is used as a parameter type in a public method. I
> > don't see why those ZipShorts cannot be converted to shorts for the
> > public API and both of them made package private.
> >
> > * ZipOutputStream: has the same problems as TarOutputStream.
> >
> > * compress.archivers.zip: every class in this package is public, I
> > don't think they all need to be.
> >
> >
> > * compress.compressors.bzip2: This package seems fine. I don't get why
> > CBZip2InputStream and CBZip2OutputStream start with the letter "C"
> > though.
> >
> >
> > * compress.exceptions: I don't get why the exceptions are segregated
> > to their own package. Javadoc already keeps them separated so there
> > isn't a need to put them in their own package. I'd put them in the
> > compress package or a package closer to where they are used.
> >
> > --
> > Sandy McArthur
> >
> > "He who dares not offend cannot be honest."
> > - Thomas Paine
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: commons-dev-help@jakarta.apache.org
> >
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.2.1 (MingW32)
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>
> iD8DBQFEMi31kv8rKBUE/T4RAhe3AJ9HCrNQf8dvHAyJnW20dWoF4tpnxwCgi7U8
> Hu/4efxAre76d8mbK/mqPrw=
> =4wUL
> -----END PGP SIGNATURE-----
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: commons-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: commons-dev-help@jakarta.apache.org
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message