Interpreting XML Tags as Type

[ernter4]

Setup
I am currently diffing a language not jet supported by Gumtree. I am currently using following Setup. At first i am parsing my language with Lark and convert the parse tree into an xml which i use gumtree to diff.

Problem
Gumtree treats the tags of xml als its own value. But this is false in this Context because the tags represent the terminals in the tree.

Proposal
Would it be possible to add an option which treats XML elements as type and its content as value.

Thanks in advance

[jrfaller]

Hi @ernter4 ! There is an official XML AST serialization format in gumtree. Could you generate your XML using this format?

Cheers.

example.xml

[ernter4]

Hi @jrfaller thanks a lot for the fast answer.
I suppose i could.
My problem is that it is not a programming language and more a description language.
Therefore the types used in the edit scripts are insufficient, because i want to change the edit scripts, and i would loose a lot of context.
I have found a workaround by diffing smaller parts of my document,
but i think this option would benefit others, who want to diff non programming languages.
Thanks a lot for your work in advance
Cheers:)

[jrfaller]

I am not sure to understand exactly what you want. Normally our XML serialization language aims at being capable at representing any asbtract syntax tree, regardless of the style of the programming language (rule-base, imperative, functionnal, etc...). Could you give an an example of what you would want that is missing?

Cheers!

[ernter4]

Hi @jrfaller , i didn't realize, that i can set the type myself and they aren't preset ....
so if i use this format and the standart xml parser it should work ?
Thx in advance:)

[jrfaller]

Exactly!

Normally, if your xml generator can be called with a command, there is a hassle free integration like this

gumtree webdiff -x lark-parser v0.mylang v1.mylang

Assuming lark-parser has the following syntax lark-parser FILE and produces the XML on the standard output.

[ernter4]

Hi thanks a lot for your support.
I am afraid to ask but is there a defenition of this format because when i try folowing script test gumtree tries to update the type 'everything' to 'something'

<?xml version="1.0" encoding="UTF-8"?>
<root>
	<context/>
	<tree type="module" pos="0" length="7492">
		<tree type="everything" label="time" pos="41" length="4"/>
	</tree>
</root>

<?xml version="1.0" encoding="UTF-8"?>
<root>
	<context/>
	<tree type="module" pos="0" length="7492">
		<tree type="something" label="time" pos="41" length="4"/>
	</tree>
</root>

sry for taking up your time :)
cheers

[ernter4]

Hi @jrfaller,
just to clarify which behavior i expect, take these two xml files:

old.xml

new.xml

One of them is the example script you gave me,
the other one has one small change, the type of the identifier in line 1548 was changed to test.
Because the type is different i expect gumtree to trigger a delete and insert action.
But when i diff the old script with the new one following action is being made:

===
update-node
---
ATTR_VALUE: "identifier" [96542,96554]
replace "identifier" by "test"
===

this is the command i used:

./gumtree textdiff /tmp/old.xml /tmp/new.xml -x "cat $1"

Is that behavior wanted ??
Thx in advance

[jrfaller]

Hi!

There is a problem with the used generator, it uses the regular XML generator instead (aiming at diffing XML files) of the internal one (aiming at serialize/deserialize parse trees). Unfortunately I don't manage to reproduce the problem since on my matching it simply does not compute the diff. I will investigate. In the meantime maybe you can try writing a specialized main to see if it works when you force the use of the deserializer (example in the tests)?

[jrfaller]

Could you try :

./gumtree textdiff /tmp/old.xml /tmp/new.xml -x "cat $FILE"