Metha-Cat: Support for Paging?

[tobiasschweizer]

Hi there,

We are using metha-sync to harvest quite a big set. Everything went smoothly and we could create an XML using metha-cat containing the whole set. The XML is quite big (2.5 GB) and we have some difficulties processing it.

Is there a way to get the records in steps of a limited size (like paging with a defined size and offset) with metha-cat? Setting the from or until params wouldn't help us much I think (libraries might process big batches on a single day).

Thanks a lot!

[miku]

I ran into similar issues (too big XML files) in the past and I remember that there are tools addressing this problem specifically; one I remember is xml_split.

On debian/ubuntu it seems to be available with the xml-twig-tools package.

$ metha-sync https://yareta.unige.ch/oai && metha-cat https://yareta.unige.ch/oai | xml_split -l 1 -s 50kb
$ ls -l
total 456K
drwxrwxr-x  2 tir tir 4.0K Jun  1 13:29 ./
drwxr-xr-x 39 tir tir  56K Jun  1 13:28 ../
-rw-rw-r--  1 tir tir  346 Jun  1 13:29 out-00.xml
-rw-rw-r--  1 tir tir  50K Jun  1 13:29 out-01.xml
-rw-rw-r--  1 tir tir  54K Jun  1 13:29 out-02.xml
-rw-rw-r--  1 tir tir  51K Jun  1 13:29 out-03.xml
-rw-rw-r--  1 tir tir  50K Jun  1 13:29 out-04.xml
-rw-rw-r--  1 tir tir  51K Jun  1 13:29 out-05.xml
-rw-rw-r--  1 tir tir  51K Jun  1 13:29 out-06.xml
-rw-rw-r--  1 tir tir  51K Jun  1 13:29 out-07.xml
-rw-rw-r--  1 tir tir  23K Jun  1 13:29 out-08.xml

The resulting XML is valid, but slightly modified:

$ xmllint --format out-01.xml 2> /dev/null | head -4
<?xml version="1.0"?>
<xml_split:root xmlns:xml_split="http://xmltwig.com/xml_split">
  <record xmlns="http://www.openarchives.org/OAI/2.0/">
    <header status="">

Does this help?

PS: Thanks for using metha! I'm just curious (and collecting uses of metha) - if possible, can you share the project name in which metha is used for data acquisition?

[tobiasschweizer]

@miku Thanks a lot for your answer. We'll try this out!

We are using metha for our linked open research data project Connectome.

[miku]

This is go-specific, so I leave this here as a footnote: I had some success making XML processing faster by parallelizing it, with some ideas take from here: Faster XML processing in Go -- anecdata: 5GB of XML can be processed in a few seconds.