Miraculous behavior of readstruct() when importing Attributes with a single d or e

Hi,

I have some question regarding readstruct(). I have a XML called test.xml which looks like this:

<folder>
    <file name="Test1.txt" CRC32="001c87e9" MD5="khcme9u7c3hgjvdrctx3u2u9l0ebj7qj" SHA1="nxnh6ywylck2rnplyl8isksn73kr8beut41jmgtv"/>
    <file name="Test2.txt" CRC32="398c7b1a" MD5="n8le51jc8elgqcczr9x6fjqvk7b243a5" SHA1="f8u0x6ftsgskel1cohdgwwz1qjb3vkmznno05ldb"/>
    
    <file name="Test3.txt" CRC32="00d02104" MD5="wx39m75pw573aj12qkwaqsc84vjsni3f" SHA1="c9svilvjwlwyo1s2k5yrh19s41z33lhc0h2wpf6o"/>
    <file name="Test4.txt" CRC32="0d392864" MD5="hoa7484gbb5b9fxhirbh56mo9x4awufj" SHA1="nnhkvy31xprai5slctk1q4urjbpx8p21lqsi7l4e"/>
    
    <file name="Test5.txt" CRC32="409558e8" MD5="slojwfowq7lqueoq35qpituuc206m050" SHA1="rcb0kulbclkfqjq7qk6yrem6zsyzvu6gnt6lh0at"/>
    <file name="Test6.txt" CRC32="52392e77" MD5="xbzxu2tphoa9zh1uk553w4xuoztln04g" SHA1="t5930b1282k094hr618pmep7hriq9n2ks2xk1fo1"/>
    
    <file name="Test7.txt" CRC32="007d4024" MD5="tzgh3ljocpvzq4rc7d1f2hfow9tg21hf" SHA1="ke4c394hr6182a2d5b7b778ke4c38ke4c30fke3p"/>
    <file name="Test8.txt" CRC32="576d9391" MD5="shi8gna8ll2vxm2fnt3g08c4yloa503s" SHA1="b3d65fhhes3e4b32fzzqlke4c30f61b3dchaz526"/>
</folder>

I want to import this as a struct into MATLAB and therefore run these two lines:

Test = readstruct("test.xml");
Test = Test.file;

What I end up with is a struct with 4 fields: nameAttribute, CRC32Attribute, MD5Attribute and SHA1Attribute, which is exaclty what I want.

But oddly there is a problem with CRC32Attribute:

I purposefully structured the test.xml into pairs to illustrate my problem:

• Test1.txt and Test2.txt have crcAttribute as wanted. It is identically to test.xml as expected and wanted
• Test3.txt and Test4.txt have crcAttribute as 0 which is not what I want.
• When looking at the XML data I noticed that it might relate to the fact that there is a leading 0 followed by a d and some other numbers with no other letters. Is d the opposite of e+10 and therefore sets the value to 0?
• Still I don't want a number I want the original string value not 0
• Test5.txt and Test6.txt have crcAttribute expanded by a power of 10 which is not what I want
• I think it relates to the previous point just that it is now e instead of d
• Test7.txt and Test8.txt have crcAttribute expanded by a power of 10 which is not what I want
• Here I have the d's again but why is it here Inf and not 0?

In a nutshell I want readstruct() to import the CRC32Attribute as it is in the test.xml as a string. No 0, no Inf or whatsoever.

Please note that I am aware that there are other ways to import this kind of data. I am just really curious how this specific problem is solved and what the cause is.

Thank you!

[Walter Roberson]

123d45 is the same number as 123e45 for MATLAB. MATLAB was inspired by FORTRAN. FORTRAN uses e to indicate single precision floating point and used d to indicate double precision floating point. They decided to make MATLAB accept either one.

That answers the question of what is happening. At the moment I do not have any ideas as to how to force a field to be a string.

[Les Beckham]

This one is pretty fascinating, if not "miraculous".

If you add "0x" in front of the CRC32 values in the xml, it will interpret all of the CRC32 fields as numeric hexadecimal numbers, which I'm assuming is the correct interpretation. If you want the string representation you could, of course, use dec2hex to convert to the string equivalent.

Interestingly, readtable interprets the first two rows of CRC32 as NaN instead of deciding that they are strings. I'm assuming that this is because tables need each of the variables (columns) to be the same type?.

Is it odd that struct arrays can have elements with the same fieldname that have different types as we see in the OP's example? Apparently not.

For example:

a(1).x = 1

a = struct with fields:
    x: 1

a(2).x = "1"

a = 1×2 struct array with fields:
    x

class(a(1).x)

ans = 'double'

class(a(2).x)

ans = 'string'

I'm not sure that makes sense to me.

[Eleftherios Bethmage]

@Les Beckham do you know of any way to add the 0x to each value? Imagine my XML would cover 10000 files. It would be impossible to do it by hand. Problem is that as soon as I import into matlab the information is already lost. So I must alter the file without loading it into the workspace right? Is this even possible?

Also if I use dec2hex it does not perserve any leading 0, which could be fixed easily but still is a bit unfortunate.

EDIT: fileread() and regexprep() did the trick for me. It is weird that I must go that route but it is a solution.

@Walter Roberson thank you for the explaination with d and e. I was wondering since the XML contains "123e45" with quotation marks and matlab still treats it as a number...

[AndresVar]

@Eleftherios Bethmage also you could use find/replace/replace-all to add 0x. dec2hex(num,min) has a second argument to add leading zeros

Answer 1

[Les Beckham]

@Eleftherios Bethmage

As @AndresVar says, just use a text editor to search and replace CRC32=" with CRC32="0x. Or, as you suggest you can do it with Matlab.

Also, the suggestion of using the second argument to dec2hex is a good one. For example:

% load the mat file containing table generated by using readtable on the
% modified xml file (added "0x" in front of the CRC32 attribute values)
% i.e., T = readtable('Test0x.xml')
% (since Answers doesn't allow xml attachments)
load Test0x.xml.mat
T.CRC32Attribute

ans = 8×1

1869801 965507866 13639940 221849700 1083529448 1379479159 8208420 1466798993

class(T.CRC32Attribute)

ans = 'uint32'

T.CRC32Attribute = dec2hex(T.CRC32Attribute, 8)

T = 8×4 table
    nameAttribute    CRC32Attribute               MD5Attribute                             SHA1Attribute               
    _____________    ______________    __________________________________    __________________________________________

     "Test1.txt"        001C87E9       "khcme9u7c3hgjvdrctx3u2u9l0ebj7qj"    "nxnh6ywylck2rnplyl8isksn73kr8beut41jmgtv"
     "Test2.txt"        398C7B1A       "n8le51jc8elgqcczr9x6fjqvk7b243a5"    "f8u0x6ftsgskel1cohdgwwz1qjb3vkmznno05ldb"
     "Test3.txt"        00D02104       "wx39m75pw573aj12qkwaqsc84vjsni3f"    "c9svilvjwlwyo1s2k5yrh19s41z33lhc0h2wpf6o"
     "Test4.txt"        0D392864       "hoa7484gbb5b9fxhirbh56mo9x4awufj"    "nnhkvy31xprai5slctk1q4urjbpx8p21lqsi7l4e"
     "Test5.txt"        409558E8       "slojwfowq7lqueoq35qpituuc206m050"    "rcb0kulbclkfqjq7qk6yrem6zsyzvu6gnt6lh0at"
     "Test6.txt"        52392E77       "xbzxu2tphoa9zh1uk553w4xuoztln04g"    "t5930b1282k094hr618pmep7hriq9n2ks2xk1fo1"
     "Test7.txt"        007D4024       "tzgh3ljocpvzq4rc7d1f2hfow9tg21hf"    "ke4c394hr6182a2d5b7b778ke4c38ke4c30fke3p"
     "Test8.txt"        576D9391       "shi8gna8ll2vxm2fnt3g08c4yloa503s"    "b3d65fhhes3e4b32fzzqlke4c30f61b3dchaz526"

class(T.CRC32Attribute)

ans = 'char'

[Eleftherios Bethmage]

Thanks this worked for me.