使用lxml修改大型xml文件

Question

Language :- Python 2.7.6

File Size :- 1.5 GB

XML Format

<myfeed>
    <product>
        <id>876543</id>
        <name>ABC</name>
        ....
     </product>

    <product>
        <id>876567</id>
        <name>DEF</name>
        ....
     </product>

    <product>
        <id>986543</id>
        <name>XYZ</name>
        ....
     </product>

I have to

A) Read all the nodes <product>

B) Delete some of these nodes ( if the <id> attribute's text is in python set()

C) Update/Alter few nodes ( if the <id> attribute's text is in python dict

D) Append/Write some new nodes

The problem is my XML file is huge ( approx 1.5 GB ). I did some research and decide to use lxml for all these purposes.

I am trying to use iterparse() with element.clear() to achieve this because it will not consume all my memory.

for event, element in etree.iterparse(big_xml_file,tag = 'product'):
        for child in element:
            if child.tag == unique_tag:
                if child.text in products_id_hash_set_to_delete: #python set()
                    #delete this element node

                else:
                    if child.text in products_dict_to_update:
                        #update this element node  
                        else:
                            print child.text
        element.clear()

Note:- I want to achieve all these 4 task in one scan of the XML file

Questions

1) Can I achieve all this in one scan of the file ?

2) If yes, how to delete and update the element nodes I am processing?

3) Should I use tree.xpath() instead ? If yes, how much memory will it consume for 1.5 GB file or does it works in same way as iterparse()

I am not very experienced in python. I am from Java background.

Answer 1

You can't edit an XML file in-place. You have to write the output to a new (temporary) file, and then replace the original file with the new file.

So the basic algorithm is:

• Loop over all elements.
• If the node is one to delete, proceed to the next element
• If the node is one to change, change its value
• Write out the node ««« This is the crucial bit you are missing
• When you are about to finish processing a node which is a parent of one of the new nodes, write out the new node, and remove it from the collection of new nodes.
• Close the output file
• Rename.

To answer the supplemental question: You need to realize that an XML file is a (long) string of characters. If you want to insert a character, you have to shuffle all the other ones up; if you want to delete a character, you have to shuffle all the other ones down. You can't do that with a file; you can't just delete a character from the middle of a file.

If you have millions of elements (and this is a real problem, not an exercise for a class), then you need to use a database. SQLite is my first thought when somebody says "database", but as Charles Duffy points out below, an XQuery database would probably be a better place to start given you already have XML. See BaseX or eXist for some open-source implementations.