ENH: Possible to add dtype/converters as arguments for pandas.read_xml() ?

[soliujing]

Is your feature request related to a problem?

I am using pandas lib to read xml for further processes, however a number of columns with leading ZERO are always converted to numbers, so I lost the original data.

Describe the solution you'd like

It would be great to add dtype/converter arguments for pandas.read_xml() to force pandas to interprete certain columns with given dtype/converters.
Just like similar IO read (read_csv, read_html, etc)

read_xml
read_csv

API breaking implications

Probably not, this argument could be optional.

Describe alternatives you've considered

Write my own code to pull data by each xml nodes, which results in very bad performance.

[mzeitlin11]

Could maybe be added to #40131, cc @ParfaitG if any thoughts

[ParfaitG]

Agreed! Good feature to add to running list. Also, read_xml passes parsed data to TextParser shared by other io readers.

[ParfaitG]

As a current workaround, consider running XSLT to quote the nodes with leading zeroes and then convert on the pandas side. If using the default lxml parser, XSLT 1.0 scripts are supported in read_xml. Below XSLT runs the standard Identity Template and encloses the text values of the zip with double quotes.

import pandas as pd

xml = \
'''<root>
     <row>
        <zip>08540</zip>
        <dat>123</dat>
     </row>
     <row>
        <zip>08628</zip>
        <dat>456</dat>
     </row>
     <row>
        <zip>27599</zip>
        <dat>789</dat>
     </row>
    </root>'''

xsl = \
'''<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform">
    <xsl:output method="xml" omit-xml-declaration="no" indent="yes"/>
    <xsl:strip-space elements="*"/>

    <!-- IDENTITY TEMPLATE TO COPY XML AS IS -->
    <xsl:template match="node()|@*">
       <xsl:copy>
         <xsl:apply-templates select="node()|@*"/>
       </xsl:copy>
    </xsl:template>
    
    <!-- ENCLOSE zip NODES WITH DOUBLE QUOTES -->
    <xsl:template match="zip">
      <xsl:copy>
        <xsl:variable name="quot">"</xsl:variable>
        <xsl:value-of select="concat($quot, text(), $quot)"/>
      </xsl:copy>
    </xsl:template>
    
</xsl:stylesheet>'''

df = (
    pd.read_xml(xml, stylesheet = xsl)
      .assign(zip = lambda x: x["zip"].str.replace('"', ''))
)

df
     zip  dat
0  08540  123
1  08628  456
2  27599  789

[soliujing]

@ParfaitG xsl is really a good idea !
I have tried and have questions on namespace.

• actually the XML file i am processing contains namespace at root node, I am not very familar with xsl, so any suggestion for match & replace ?

I have done as below but cannot get any output..

xsl = \
'''<xsl:stylesheet version="1.0" xmlns:xsl="http://www.w3.org/1999/XSL/Transform" xmlns:doc="http://sample.com/sample">
    <xsl:output method="xml" omit-xml-declaration="no" indent="yes"/>
    <xsl:strip-space elements="*"/>

    <!-- IDENTITY TEMPLATE TO COPY XML AS IS -->
    <xsl:template match="node()|@*">
       <xsl:copy>
         <xsl:apply-templates select="node()|@*"/>
       </xsl:copy>
    </xsl:template>
    
    <!-- ENCLOSE zip NODES WITH DOUBLE QUOTES -->
    <xsl:template match="doc:zip">
      <xsl:copy>
        <xsl:variable name="quot">"</xsl:variable>
        <xsl:value-of select="concat($quot, text(), $quot)"/>
      </xsl:copy>
    </xsl:template>
    
</xsl:stylesheet>'''

[ParfaitG]

Hmmmm...At the very least, you should get back the same XML with this XSLT. The identity transform should work across any XML with or without namespaces. I can understand the double quotes not rendering. Possibly, doc: is not the correct namespace for <zip> nodes. Please share sample of XML for reproducible example.