Hi,
On Fri, 2006-03-31 at 00:08 -0500, Kevin Kubasik wrote:
> Let me know what you all think. This is my first shot, hope it works
> out. I looked into using one of Mono's decompression libs to extract the
> actual description information, but .deb's are in the somewhat obscure
> ar archive format. So dpkg-deb it was. If anyone can think of a property
> I missed, please let me know, they are quite easy to add.
Can you attach some sample dpkg-deb output? I don't have a Debian or
Ubuntu box, so it's hard to see if we're doing the right thing with
regards to parsing.
> Also, is there a major objection to just adding the Description field as
> text? and furthermore, is it ok that this happens in DoPullProperties()?
> as opposed to DoPull()?
That's fine.
> public FilterDeb (){
> AddSupportedFlavor (FilterFlavor.NewFromMimeType ("application/x-deb"));
> AddSupportedFlavor (FilterFlavor.NewFromExtension(".deb") );
> }
Are both of these necessary? I prefer to avoid setting NewFromExtension
unless there's a chance that the mime type detection won't work without
it.
> StreamReader pout = pc.StandardOutput;
> string str = null;
> string[] tokens = null;
> string strMetaTag = null;
> bool bKeyword = false;
> bool listWords = false;
> char[] splits = {',','|'};
> string[] list = null;
> while ((str = pout.ReadLine ()) != null) {
> bKeyword = false;
> listWords = false;
> strMetaTag = null;
> tokens = str.Split (':');
> if (tokens.Length > 1) {
> switch (tokens[0].Trim()) {
> case "Package":
> strMetaTag = "dc:title";
> break;
> case "Maintainer":
> strMetaTag = "dc:author";
> break;
> case "Version":
> strMetaTag = "fixme:version";
> bKeyword = true;
> break;
> case "Section":
> strMetaTag = "fixme:section";
> bKeyword = true;
> break;
> case "Architecture":
> strMetaTag = "fixme:arch";
> bKeyword = true;
> break;
> case "Depends":
> listWords = true;
> list = tokens[1].Split(splits);
> foreach(string s in list){
> AddProperty (Beagle.Property.NewKeyword ("fixme:dep",
> s.Trim()));
> }
> break;
> case "Recommends":
> listWords = true;
> list = tokens[1].Split(splits);
> foreach(string s in list){
> AddProperty (Beagle.Property.NewKeyword ("fixme:recommend",
> s.Trim()));
> }
> break;
> case "Conflicts":
> listWords = true;
> list = tokens[1].Split(splits);
> foreach(string s in list){
> AddProperty (Beagle.Property.NewKeyword ("fixme:conflict",
> s.Trim()));
> }
> break;
> case "Replaces":
> listWords = true;
> list = tokens[1].Split(splits);
> foreach(string s in list){
> AddProperty (Beagle.Property.NewKeyword ("fixme:replace",
> s.Trim()));
> }
> break;
> case "Provides":
> listWords = true;
> list = tokens[1].Split(splits);
> foreach(string s in list){
> AddProperty (Beagle.Property.NewKeyword ("fixme:provide",
> s.Trim()));
> }
> break;
> case "Installed-Size":
> strMetaTag = "fixme:size";
> bKeyword= true;
> break;
> case "Description":
> listWords=true;
> AppendText(tokens[1]);
> AppendStructuralBreak ();
> while ((str = pout.ReadLine ()) != null) {
> if(str.Trim()==".")
> AppendStructuralBreak();
> else
> AppendText(str);
> }
> break;
> }
> if (strMetaTag != null) {
> if(!listWords){
> if (bKeyword){
> AddProperty (Beagle.Property.NewKeyword (strMetaTag,
> tokens[1].Trim()));
> }
> else{
> AddProperty (Beagle.Property.New (strMetaTag,
> tokens[1].Trim()));
> }
> }
> }
Sorry for quoting such a big block. A few things here. First, the
entire code block is within a "if (tokens.Length > 1)" block. We can
remove an entire level of indentation by just saying "if (tokens.Length
<= 1) continue;" here, assuming it's not an error.
Also, rather than setting strMetaTag, bKeyword, listWords and a handfull
of other variables and carrying that state through, I think it'd be lot
easier to just add a AddProp() and AddKeyword() method and just call
those when you need to. Method overhead is negligible. And that
removes the nasty special case you do for listWords anyway.
Joe