Vala libraries and make distcheck

GNOME Builder have support to create Vala libraries skeleton using templates, but can’t produce any useful library yet. You can see bugs 766621, 768541, 768740, 768540 and 766621, for missing features.

While my patches to improve Builder’s template system work, they can be used when you try to run make distcheck, due to current Vala support implementation of Automake. See bug GNU bug #13002.

But now I on stub! No one have checked 13002, Christian Hergert  have no time to review my patches, well not all of them and Builder is a big project, and personally I now I’m not good enough on Autotools to create a clean implementation, but with Christian feedback I’ve get my best and will write better cleaner implementation.

Is better to fix Automake Vala implementation, in order to produce clean and easy to write Makefile.am for Vala projects, less code and less work around. All implementation should pass both make check and make distcheck.

If no response found, then may I need to write Autotools macros and mk files to be imported to your Makefile.am file in order to provide a clean implementation and easy to use, only after Autotools is fixed.

This will me help too, to write my own libraries more easy. Now I have to fix manually and find the current configuration for each one, most of the time based on GXml, but is a time consuming task, requires Autotools knowlage and should be fixed each time you added new stuff. I really like libgee’s Makefile.am, which is clean and just works, but all the time I start to use it as template, it just not work for me. Then I think I need to create my own and push it to Builder to produce good templates for all of you too.

Removing Gee warnings

If you have been using Gee for a while, you have noticed it produce a lot of warnings not at Vala compilation but at C compilation time.

GXml is using Gee for a while too and we are facing the same. I have some proprietary applications using Gee too.

Recently I started to search the source of these warnings. I found valac producing C code with simple cast to parent but not the  current class, this is:

_tmp25_ = gee_abstract_collection_get_size ((GeeCollection*) _tmp24_);

This function requires to cast to GeeAbstractCollection, this is why we see some of these warnings.

In order to reduce this warnings I started to use “as” operator. When transformed in C, it first check if the object to cast is the type of required one, if not it just pass a NULL pointer, on success it just cast.

Now in my internal code, and soon on GXml, I have used “as” operator, to reduce this C warnings and producing less stress. But at the same time, when you have a C warning, you should fix it, in order to reduce the risk of improper behavior or crash. Now most of you using Gee or GXml, have lot of warnings and you have chosen to ignore them, because is not produced by your code; may in some times is the case, but when you see lot of them, you prefer to ignore all if valac no warns you.

I’m sure we need to change some Vala code at Gee source code level, but that is another history and I’ll go for them, when have time enough; if you can help, please file bugs with patchs and I can help you to push to Gee or GXml, making Vala programming a pleasure as is today.

EDIT:

Ok, it is unexpected, I have to admit, but Gee has a clean compilation without any warning.

At the same time, while changing my own code, found changing simple properties like size to (object as Gee.Map).size is enough to remove annoying warnings, but to get objects on foreach statement you should change from:

foreach (MyObject o in collection_object.values)

to

foreach (MyObject o in (collection_object as Gee.Map<string,CollectionObject>).values)

requiring more work at writing your code time.

Conclusion: This is a valac bug, producing C code without correct casts.

DOM4 API and GXml API

Now while fixing compilation of GXml’s implementation of DOM4, I found lot of API incompatibilities between existing generic GXml API using Gee collections and DOM4 interfaces.

In some cases when API incompatibilities comes from methods names, no problem, because just instruct Valac to override the correct interface, but properties is another beast.

Properties should be same type between interfaces if you want to implement one property with the same name in two or more interfaces.

While I see lot of advantages to use DOM4 API, because it is well documented in its specification, I’m using it when conflicts appear, modifying existing one in GXml.

Doing so and while GXml is perfectly usable in many cases, like embedded in Python code for web pages, DOM4’s API is more binding friendly than Vala’s generic one, while the last is powerful and easy to use in Vala code.

All this work will be present in upcoming GXml 0.11 version, for you to test. Next versions of 0.11.x series will be to fix DOM4 implementations by adding Unit Tests.

GXML and DOM4

Today I’ve finished to prototype most DOM4 interfaces implementation on GNode series.

I’m starting to fix building and some issues here and there. All my work is in my local repository jet, just want to provide a buildable commit to you, releasing a 0.11.0 development release.

DOM4 makes some changes, making old DOM implementation in GXml, well, obsolete. Once I’ve finished to add all Unit Test and passing, I think I can promote a new release 0.12.

May be at 0.14 release window and with user’s feedback on DOM4 implementation, I can remove old DOM implementation and all xNode classes, by cleaner, faster and more maintainable GNode classes.

There are lot of interfaces in DOM4 I have no implementations in GXml, just leave there to implement later. I hope we can have a clean interfaces before to implement in order to avoid API changes, but any way, this is why we are at 0.x release series.

Using GNOME Builder for Vala development

Vala programing language is, let say, misunderstood most of the time. May this explain, why there are few support for IDEs, like GNOME Builder, at least at beginning.

In 3.20.4, GNOME Builder has improved and stabilized its support for Vala development, proven to be stable enough for daily usage. Even I started to use it for GXml upstream development. But it was a little difficult, because I want to stay in a stable long term supported OS, like Debian 8.

Debian 8, has been discarded by GNOME wiki, to install and develop using jhbuild, making hard for me to develop upstream. In order to solve this, I started to test Builder from xdg-app and nightly builds, because, it suppose to be in sync with master (upstream) versions of basic Platform, for upcoming 3.22. While I rather happy with Builder running this way, you need some extra configuration to develop any Vala project under xdg-app sanboxed environment.

While Vala 0.32 is available in org.gnome.Sdk 3.20, the one I’m using to develop GXml, no libgee is installed and you can’t simple compile and install it, because in the Builder sanboxed environment, /usr is mounted read-only. So I first:

  1. Get libgee sources
  2. Set “Installation Prefix” to /home/yourusername/.local
  3. Disable instrospection (–disable-introspection), by setting “Configure Options”
  4. Use “Host operating system”

This will make the trick to compile and install libgee in your home directory. Libgee doesn’t requires other dependencies. But for GXml, we need a few more work, because you should setup it to find pc files for libgee and more important, to find its VAPI file.

To configure GXml on Builder, in sanboxed environment:

  1. Get sources
  2. Set “Installation Prefix” to /home/yourusername/.local
  3. Disable instrospection (–disable-introspection), by setting “Configure Options”
  4. Use “Host operating system”
  5. Set “Environment” variables for VALAFLAGS and PKG_CONFIG_PATH
  6. In your Makefile.am make sure you add –vapidir=/home/yourusername/.local/share/vala-0.32/vapi to your Vala compile flags.

VALAFLAGS is useful for code completion in Builder, add most of your project Vala flags. PKG_CONFIG_PATH should be set to /home/yourusername/.local/lib/pkgconfig in order to find pc files of your locally installed dependencies.

You’ll find Builder complaint by reporting some warnings, most of them coming from libgee usage.

Building your Vala project in Builder, is no jet really integrated, because most errors are shown in terminal “Build Output”, but not in build messages, no other than the libgee ones. Fortunately, you have color output from valac, making more easy to find errors and warnings.

Most of you may will use a distribution, to use Builder and jhbuild, making easy to develop upstream. I can’t do that, because some other activities requires me to have an stable system, avoid interruptions because my unpaid contributions to GNOME.

I’ve started to add DOM 4 API support for GXml

Using W3C specifications for DOM 4, I’ve managed to copy & paste its interfaces, then with few adoptions are now GObject interfaces thanks to Vala.

Now I’m implementing them using GNode classes, this is in order to deprecate xNode classes, while we provide a more modern API.

The progress are so straightforward, at least for now, with few code. I’ll try to avoid any API break this time, unless it is impossible to implement DOM4 API.

While actual API is powerful enough to do most DOM4, there are missing or just useful API we can take advantage of.

For example, DOM4 provides events API, witch we can implement provide using GObject signals, but even so I’ll try to do my best to keep the DOM’s one.

I will implement step by step each interfaces functionality, on existing classes; may I leave events for a second try. I need to implement all basic interfaces, before I can push to repository.

Even so, all API interfaces are in place, and may will suffer some Vala adaptions, but not too much. If I can’t implement them I’ll hide them from public API, until they are implemented.

HowTo: GObject Serialization to XML using GXml

While GXml has C API, it is written in Vala, then this posts will show you how to use GXml’s Vala API to get GObject to XML and back easy.

Lets start with a simple class and GXml’s implementation of GXml.Serializable, called GXml.SerializableObjectModel:

using GXml;

public class Record : GXml.SerializableObjectModel {
  public string name { get; set; }

  // GXml.SerializableObjectModel overrides
  public override string node_name () { return "Record"; }
  public override string to_string () { return @"Record: $name"; }
}

The above example, defines a class called Record witch should be written in XML as:

<Record name="Daniel"/>

This just happen if Record.name is set, if no you’ll get:

<Record/>

In order to get this XML to be written you’ll need:

var r = new Record ();
r.name = "Daniel";
var d = new GDocument ();
r.serialize (d);
stdout.printf (@"$d");

The above code will print an XML representation of your object, with the given values, as shown above.

If you have an XML file or string, you can use following to read data from:

var r = new Record ();
var d = new GDocument.from_string ("<Record name=\"Jhon\"/>");
r.deserialize (d);
stdout.printf (@"$r");

The above code will print your object string representation using Record.to_string().

GXml.SerializableObjectModel will inspect all your public properties, trying to print out string representation of each. Most basic types are sopported, while you can override default GXml.SerializableObjectModel.serialize_property() in order to catch up when an special property like GLib.List is corrently serialize to create your custom serialization; if you do so, remember to override GXml.SerializableObjectModel.deserialize_property() to get back string to your property. Remember, GXml.SerializableObjectModel is just an implementation of GXml.Serializable interface and you can both, override default methods or implement your own.

Next time we can review how a set of XML tags could be de/serialized from XML using builtin GXml serializable collections.

XML GObject Serialization

GXml is a library, written in Vala, providing a good support to read and write XML files. Behind it you’ll find libxml2, wrapped by GObject classes.

Initially written by Richard Schwarting, to provide DOM Level 3 support on libxml2, now it has a set of interfaces, making easy to write different backends. While today just libxm2 is used, GXml have three GObject classes, one in with the original DOM support, one optimized to gain better use of libmxl2 and one in pure GObject. All of them allows you to access XML objects, like tags and properties, navigate over XML tree and modify if you need it. Each implementation of general purpose interfaces, like GXml.Node, have its own advantanges: DOM support, low memory foot print or better performance with larger foot print. Is a matter of you on what implementation you need.

My motivation on GXml, was initially pushed to have a low C library to support reading Substation Configuration Language files, IEC 61850-6 standard XML based files to energy related devices configuration. This files are large and complex. While there are a free open source project to handle them by LibreSCL, in order to have it working, GXml should gain GObject to XML serialization/deserialization. This has lead to commercial products, running both on Linux and Windows.

I’ll write down about how GXml serialization can be used for GObject Serialization, wait second posts.

GDA becomming better introspectable

GNOME Data Access (GDA), is a C library for Database access, it provides a library to write clients and a set GTK+ widgets to help write GUI applications; it also provides a control center, a SQL browser and CLI a la PostgreSQL’s psql command.

It is really useful, but could be better if most of its advanced futures are available for GObject Introspection (GI). While GDA has support for GI for a while, there are some structures in C and some API witch has been marked as no-introspectable, automatically or hard coded.

Making a C library introspectable is a little easy, but requires a lot of work, by hiding structs members and provide getters and setters methods for its members (sealing); make sure you are handling correctly struct’s memory management, freeing on request, but most importantly, is to convert this struct as a GBoxed type, making easy for bindings to create, read/write members (using its setter/getter), free, copy  and pack it to a GValue.

The hard work on making a struct introspectable, is to substituted all direct access to its members to use its new API. This is a hard work as you can see when GdaBinary was sealed to make it introspectable.

You can follow GDA Instrospectable bug and its bug dependencies, in order to know how has been solved each case; may that will help you to create better C introspectable API or help us in the process.