
Hi guys. I've got a folder with about 80 XML files in it. I want to take each file and make specific modifications to it. (Mostly just finding specific attributes and changing their values to make then all consistent.) Now I guess it wouldn't take me /that/ long to code something from scratch. But does anybody have a better suggestion for tools or libraries that might be useful?

Andrew Coppin
I've got a folder with about 80 XML files in it. I want to take each file and make specific modifications to it. (Mostly just finding specific attributes and changing their values to make then all consistent.)
Now I guess it wouldn't take me /that/ long to code something from scratch. But does anybody have a better suggestion for tools or libraries that might be useful?
HaXml Before google earth exposed a facility for computing path lengths, I wrote my own using HaXml. The guts of it look like this: main … xml <- fmap (xmlParse filename) (readFile filename) print_path_length xml print_path_length (Document _ _ content _) = print $ compute_length $ coordinate_string_to_list coordinate_string where coordinate_string = singleCString . (txt `o` children `o` deep (tag "coordinates")) $ (CElem content) which gives some idea of the flavour. -- Jón Fairbairn Jon.Fairbairn@cl.cam.ac.uk http://www.chaos.org.uk/~jf/Stuff-I-dont-want.html (updated 2010-09-14)

On 23/11/2011 10:14 AM, Jon Fairbairn wrote:
Andrew Coppin
writes: I've got a folder with about 80 XML files in it. I want to take each file and make specific modifications to it.
HaXml
Mmm. That looks very promising...
which gives some idea of the flavour.
OK. So it looks like processXmlWith is the function I want, if I'm going to read one file and create another from it. So now I just need to figure out which combinators I need. (The documentation seems a bit thin.) Can you show me a snippet for how I would find [one] element named "foo" and change its "bar" attribute to have the value "5"?

On 23/11/2011 12:58 PM, Andrew Coppin wrote:
On 23/11/2011 10:14 AM, Jon Fairbairn wrote:
HaXml
Mmm. That looks very promising...
which gives some idea of the flavour.
OK. So it looks like processXmlWith is the function I want, if I'm going to read one file and create another from it. So now I just need to figure out which combinators I need. (The documentation seems a bit thin.) Can you show me a snippet for how I would find [one] element named "foo" and change its "bar" attribute to have the value "5"?
Well, from what I've been able to gather, HaXml has a really nice filter combinator library. However... Weird thing #1: processXmlWith handles the common case of loading a file from disk, filtering it, and saving the result to disk again. However, it does this based on CLI arguments. There is no function anywhere that I can find which allows the host program to specify what files to process. If you want to do that, you have to reimplement most of the body of this function all over again yourself. That seems a strange omission. Weird thing #2: There are absolutely no filters for dealing with attributes. I couldn't find anything anywhere that says "apply this function to all the attributes of this element". I can find a function to /replace/ an element's attributes without regard to what existed before. But even something as trivial as adding an additional attribute while keeping the existing ones doesn't appear to be supported at all. Fortunately it turns out to not be especially hard to read the source for the replace-attributes function and change it to do what I want. But, again, it seems a rather large and obvious ommission. (I'm guessing that since attributes are key/value pairs and not "content", you would need a seperate "attribute filter" type, which is different from the existing content filters. Even so, it shouldn't be /that/ hard to do...) Anyway, the important thing is, Haskell (and more specifically HaXml) let me accomplish the task I wanted without too much fuss. It's /certainly/ faster than editing 80 files by hand in a text editor!

Andrew Coppin
On 23/11/2011 12:58 PM, Andrew Coppin wrote:
On 23/11/2011 10:14 AM, Jon Fairbairn wrote:
HaXml
Mmm. That looks very promising...
which gives some idea of the flavour.
OK. So it looks like processXmlWith is the function I want, if I'm going to read one file and create another from it. So now I just need to figure out which combinators I need. (The documentation seems a bit thin.) Can you show me a snippet for how I would find [one] element named "foo" and change its "bar" attribute to have the value "5"?
Well, from what I've been able to gather, HaXml has a really nice filter combinator library. However...
Weird thing #1: processXmlWith handles the common case of loading a file from disk, filtering it, and saving the result to disk again. However, it does this based on CLI arguments. There is no function anywhere that I can find which allows the host program to specify what files to process. If you want to do that, you have to reimplement most of the body of this function all over again yourself. That seems a strange omission.
Weird thing #2: There are absolutely no filters for dealing with attributes. I couldn't find anything anywhere that says "apply this function to all the attributes of this element". I can find a function to /replace/ an element's attributes without regard to what existed before. But even something as trivial as adding an additional attribute while keeping the existing ones doesn't appear to be supported at all.
Fortunately it turns out to not be especially hard to read the source for the replace-attributes function and change it to do what I want. But, again, it seems a rather large and obvious ommission. (I'm guessing that since attributes are key/value pairs and not "content", you would need a seperate "attribute filter" type, which is different from the existing content filters. Even so, it shouldn't be /that/ hard to do...)
Anyway, the important thing is, Haskell (and more specifically HaXml) let me accomplish the task I wanted without too much fuss. It's /certainly/ faster than editing 80 files by hand in a text editor!
I think these observations should be addressed to Malcolm Wallace. -- Jón Fairbairn Jon.Fairbairn@cl.cam.ac.uk
participants (2)
-
Andrew Coppin
-
Jon Fairbairn