I don't quite understand how to use application specific YAML tags, and maybe its because my desired use of them is purely wrong. I am using YAML for a configuration file and was hoping to use tags to provide my configuration loader with a hint as to what datatype it should parse the data into - application specific datatypes.
I'm also using libyaml with C.
So I'm trying to do something like...
shapes:
square: "0,4,8,16"
circle: "5,10"
In my app I'd like to use tags as hints so I can load the values of square into my square data structure, and the values of circle into my circle data structure (these values mean nothing in this example).
So I'm currently doing:
shapes:
square: !square "0,4,8,16"
circle: !circle "5,10"
Libyaml will provide a tag of "!square" when I'm passed the scalar "0,4,8,16". Is it valid to use this tag to provide my loader with a hint of how to process the scalar?
Since it does work for me, I'm more curious to know if its proper. And if not, how would I go about making this more proper.
Thanks.
I know that this is an ancient question, but anyway I've seen !int, etc being used in yaml files before so I went to look up the specs at Yaml 1.2 Spec # Tags
application specific tag: !something |
The semantics of the tag
above may be different for
different documents.
As per the document, it does look like your intended usage of tags is correct for application specific tag.
Related
I'm writing MapBox vector tiles using geotrellis vectorpipe.
see here for the basic flow: https://geotrellis.github.io/vectorpipe/usage.html
Typically GeoJson Features can have an id field, so that Features can be rolled up into FeatureCollections. I need to make use of this field, but vectorpipe doesn't (natively) have this capability.
This is the Feature type used, and you can see it only has space for 1) a Geometry and 2) a data object D (which ends up populating properties in the output). There is no spot for an id.
https://geotrellis.github.io/scaladocs/latest/index.html#geotrellis.vector.Feature
Upstream there's a method called writeFeatureJsonWithID() that does let you inject an id field into a Feature when writing GeoJson.
https://github.com/locationtech/geotrellis/blob/master/vector/src/main/scala/geotrellis/vector/io/json/FeatureFormats.scala#L41-L49
My question is this:
I have worked through the vectorpipe code (https://github.com/geotrellis/vectorpipe), and I can't figure out if/where the data ever exists as GeoJson in a way where I can override and inject the id, maybe using the writeFeatureJsonWithID() or something I write explicitly. A lot of conversions are implicit, but it also may never explicitly sit as json.
Any ideas for how to get an id field in the final GeoJson written to vector tiles?
EDIT
Right now I think the trick is going to be finding a way to override .unfeature() method here:
https://github.com/locationtech/geotrellis/blob/master/vectortile/src/main/scala/geotrellis/vectortile/Layer.scala
The problem is that the internal.vector_tile.Tile is private, so I can construct it without forking the project.
Ended up having to fork geotrellis, hard-code a metadata => id function in Layer.unfeature() and compile locally to include in my project. Not ideal, but it works fine.
Also opened an issue here: https://github.com/locationtech/geotrellis/issues/2884
I noticed that hyperHTML preserves references I make to elements:
let div = document.createElement("div");
div.textContent = "Before Update";
hyperHTML.bind(document.body)`static1 - ${div} - static2`;
div.textContent = "After Update";
Above will produce a page that says:
static1 - After Update - static2
It is my understanding that hyperHTML ultimately clones an HTML <tempate> element to render the final output. However, don't you typical lose references when cloning an HTML template (like the variable "div" in the example above)?
Therefore, on the initial render, does hyperHTML somehow replace cloned elements with their originals after cloning the HTML template?
Here's how I think it works:
Create an HTML Template of the original template literal while
replacing all interpolations with comments.
Clone the html template with comments left in.
Make elements or document fragments out of each interpolation originally recieved
Replace each comment in the clone with its processed interpolation.
Is this correct?
I am not sure what is the question here, but there is a documentation page, and various examples too to understand how to use hyperHTML, which is not exactly in the way you are using it.
In fact, there's no need to have any reference there because hyperHTML is declarative, so you'd rather write:
function update(text) {
var render = hyperHTML.bind(document.body);
render`static1 - <div>${text}</div> - static2`;
}
and call update("any text") any time you need.
Here's how I think it works ... Is this correct?
No, it's not. hyperHTML doesn't clone anything the way you described, it associates once per unique template tag a sanitized version to the output and finds out all interpolated holes in it.
The part of the library that does this is called domtagger, and the mapping per template literal is based on the standard fact that these are unique per scope:
const templates = [];
function addTemplate(template, value) {
templates.push(template);
return template.join(value);
}
function asTemplate(value) {
return addTemplate`number ${value}!`;
}
asTemplate(1);
asTemplate(2);
asTemplate(Math.random());
templates[0] === templates[1]; // true
templates[1] === templates[2]; // true
// it is always the same template object!
After that, any other element using once that very same tag template will have a clone of that fragment with a map to find holes once and some complex logic to avoid replacing anything that's already known, being that text, attributes, events, or any other kind of node.
hyperHTML never removes comments, it uses these as pin and then uses domdiff to eventually update nodes related to these pins whenever there's a need to update anything.
Domdiff is a vDOM-less implementation of the petit-dom algorithm, which in turns is based on E.W Myers' "An O(ND) Difference Algorithm and Its Variations" paper.
Whenever you have DOM nodes in the holes, hyperHTML understand that and fill these holes with those nodes. If you pass repeatedly the same node, hyperHTML won't do anything 'cause it's full of algorithm and smart decisions, all described in the documentation, to obtain best performance out of its abstraction.
All these things, and much more, normalized for any browser out there, makes hyperHTML weight roughly 7K once minified and gzipped, bit it also offers:
Custom Elements like hooks through onconnected/disconnected listeners
lightweight components through hyperHTML.Component
SVG manipulation as content or via wire
easy Custom Elements definition through HyperHTMLElement class
As summary, if you need these simplifications and you don't want to reinvent the wheel, I suggest you give it a better try.
If you instead are just trying to understand how it works, there's no need to assume anything because the project is fully open source.
So far, all I've read from your questions here and there, is that you just believe to understand how it works so I hope in this reply I've put together all the missing pieces you need to fully understand it.
Do you want to write your own lit/hyperHTML library? Go ahead, feel free to use the domtagger or the domdiff library too, few others are already doing the same.
I have a need to go through a set of DICOM files and modify certain tags to be current with the data maintained in the database of an external system. I am looking to use GDCM. I am new to GDCM. A search through stack overflow posts demonstrates that the anonymizer class can be used to change tag values.
Generating a simple CT DICOM image using GDCM
My question is if this is the best use of the GDCM API or if there is a better approach for changing the values of individual tags such as patient name or accession number. I am unfamiliar with all of the API options but have a link to the API documentation. It looks like the DataElement SetValue member could be used, but it doesn't appear that there is a valid constructor for doing this in the Value class. Any assistance would appreciated. This is my current approach:
Anonymizer anon = new Anonymizer();
anon.SetFile(myFile);
anon.Replace(new Tag(0x0010, 0x0010), "BUGS^BUNNY");
Quite late, but maybe it would be still useful. You have not mention if you write in C++ or C#, but I assume the latter, as you do not use pointers. Generally, your approach is correct (unless you use System.IO.File instead of gdcm.File). The value (second parameter of Replace function) has to be a plain string so no special constructor is needed. You should probably start with doxygen documentation of gdcm, and there is especially one complete example. It is in C++, but there should be no problems with translation.
There are two different ways to pad dicom tags:
Anonymizer
gdcm::Anonymizer anon;
anon.SetFile(file);
anon.Replace(gdcm::Tag(0x0002, 0x0013), "Implementation Version Name");
//Implementation Version Name
DatsElement
gdcm::Attribute<0x0018, 0x0088> ss;
ss.SetValue(10.0);
ds.Insert(ss.GetAsDataElement());
I've got a setup where I use doxygen to describe a set on unit tests (I use QtTest to run the tests). The output from the tests are parsed by a little Python snippet that produces a nice and tidy report. Now, I'd love to link from the report to each test case, i.e. private slot member method, in the doxygen material. However, the anchors defined by doxygen looks like this:
<a class="anchor" id="a2a0e066d4dad8e0dff6c9231bf65fd65"></a>
<!-- doxytag: member="PRadioTunerTst::scanFM" ref="a2a0e066d4dad8e0dff6c9231bf65fd65" args="()" -->
Sure, I could parse the doxygen html and match all method to the reference key, but I'd much rather have readable links. I do not overload any unit test case methods, so having them enumerated would not be an issue - I'd simply be able to pick the first and only. I'd even be happy to calculate the id hash myself. I just need to know how to.
So, basically, the questions is:
Does anyone know how to tune doxygen
to generate readable anchors
if not, how do I calculate the hash?
Instead of trying to reconstruct the hash (which is a md5 checksum over the method's definition as parsed by doxygen, see MemberDef::setAnchor() in the code). I would suggest to let doxygen generate a tag file (GENERATE_TAGFILE) and then parse that. The tag file is a simple XML file which has both the name and the anchor for each member.
I also needed link targets, in my case for rst docs to point at breathe/doxygen-created html. Here's what I did:
To better understand how doxygen creates anchors, I recompiled doxygen, with this at the end of setAnchor():
printf("memAnchor=%s sigStr=%s\n", memAnchor.data(), sigStr.data());
, which creates output like:
memAnchor=const int SomeNamespace::GetStateGetState(SomeNamespace::State *state) sigStr=f2c41a8a6a152602c92fefb80bd0862a
I already had my function signatures, so I created strings similar to memAnchor above and piped it through md5sum to get the hash, then appended it to the string which is common to all anchors. In my rst doc, I then put definitions like:
.. _GetState: `project0class_SomeNamespace_1f2c41a8a6a152602c92fefb80bd0862a`_
Not sure on the first question about readable anchors.
I posted a question on the DOxygen forums and also am posting it here for a better response.
I have a moderately sized C project of about 2,900 functions. I am using DOxygen 1.5.9 and it is successfully generating a call graph for the functions. Is there a way to extract this out for further analysis? A simple paired list would be sufficient, e.g.
Caller,Callee
FunctionX, FunctionY
...
I am comfortable with XSLT but I must say that the DOxygen XML output is complex. Has anyone done this before and can provide some guidance on how to parse the XML files?
Thanks in advance!
Based on what I see in the contrived example that I created,
Parse files with a name similar to ^_(.+)\d+(c|cpp|h|hpp)\.xml$, if my regex-foo is right.
Find all <memberdef kind="function">. It has a unique id attribute. I believe the XPath for this is //memberdef[#kind='function'].
Within that element, find all <references>.
For each of those tags, the refid attribute uniquely refers to the id attribute of the corresponding <memberdef> that is being called.
The text node within each <references> corresponds to the <name> of the corresponding <memberdef> that is being called.
This seems like a nice, straightforward way to express call graphs. You should have no trouble using XSLT or any other sane XML-parsing suite to get the desired results.