Can you demonstrate this:
>> Sure, it broke detection and resulted in a "null" device before the
>>incomplete "alias" was put into DeviceData incorrectly.
My understanding was that nothing was broken. As I pointed out before, you
incorrectly ran the unit tests created for 1.0.1 against a 1.0.2 snapshot.
Thats it. Since the 1.1.0 client was just released and 1.0.2 data is next, I
will be updating the 1.1.0 unit tests to work against 1.0.2 and eventually
1.0.3. It has to be done this way because we need reproducible tests, ie tests
need to be pinned to data versions, and we cannot release software with
snapshot dependencies. So if someone downloads the 1.1.0 data release, it will
run tests against 1.0.1, validating the release as we released it. Make sense?
---
From: Werner Keil <[email protected]>
To: [email protected]; Reza Naghibi <[email protected]>
Sent: Friday, January 2, 2015 4:38 PM
Subject: Re: Improper "aliasing" of devices
Sure, it broke detection and resulted in a "null" device before the incomplete
"alias" was put into DeviceData incorrectly.
After the fix both SGP311 and SGP312 (or future UA variations like SGP313) UA
are properly detected. Before neither the valid SGP311 nor the incomplete
SGP312 were recognized. Sorry but since BuilderData source contains important
patterns for thousands of devices the classifier has to stick to the correct
format of these files as with all the other entries.
Should you have a completely separate specs for SGP312 other than <device
id="SGP311" parentId="genericSonyEricsson"> <property name="model"
value="SGP311"/> <property name="marketing_name" value="Xperia
Tablet Z"/> <property name="displayWidth" value="1200"/>
<property name="displayHeight" value="1920"/> <property
name="mobile_browser" value="Android Webkit"/> <property
name="device_os" value="Android"/> <property
name="device_os_version" value="4.1.2"/> <property
name="dual_orientation" value="true"/> <property name="inputDevices"
value="touchscreen"/> <property name="is_tablet" value="true"/>
<property name="ajax_support_javascript" value="true"/>
<property name="ajax_support_getelementbyid" value="true"/>
<property name="ajax_support_inner_html" value="true"/> <property
name="ajax_manipulate_dom" value="true"/> <property
name="ajax_manipulate_css" value="true"/> <property
name="ajax_support_events" value="true"/> <property
name="ajax_support_event_listener" value="true"/> <property
name="image_inlining" value="true"/> <property name="from"
value="devicemap"/> </device>
please feel free to add a second independent <device> entry.
DeviceData is full of Thousands of these, see various Droid XYZ, etc.If parent
was legitimate for all those it would have been done, but it isn't.
Use <device id="SGP311"> <list>
<value>SGP31\d</value> </list>
or something like <device id="SGP311"> <list>
<value>SGP311</value> <value>SGP312</value>
</list>in BuilderDataSource for model aliasing.
Thanks,Werner
On Fri, Jan 2, 2015 at 10:26 PM, Reza Naghibi <[email protected]>
wrote:
So you cannot demonstrate the client breaking on the parent. Can you please
revert you commits and restore the changes requested by Infomaker Scandinavia
AB on DMAP-86 [0]?
[0] https://issues.apache.org/jira/browse/DMAP-86
---
From: Werner Keil <[email protected]>
To: [email protected]; Reza Naghibi <[email protected]>
Sent: Friday, January 2, 2015 4:11 PM
Subject: Re: Improper "aliasing" of devices
For certain things like "parent devices" yes, but as I mentioned earlier (and
unlike the incomplete device entry "leaf" this doesn't cause a mal-detection)
the W3C DDR implementation is much more data-driven here than Classfier having
just a few patterns hard-coded in a single Java class.
BuilderDataSource has been full of device-specific regex patterns
like"[Bb]lack.?[Bb]erry" making it flexible to a wide range of matching User
Agents.
I could not find evidence any of that is used by Classifier, instead only the
class name of the branch is split into "simple" or "android", etc. assuming
these values are just keys for the other DeviceData source, which is not
entirely correct. The regEx and even concrete strings are matched agains
relevant sections of the UA, not the deviceId, which is why an exact same
pattern in the "Droid X 2" case may have got a vague result here, but it didn't
break anything and the duplicate was also around since 2012 as of OpenDDR.
Werner
On Fri, Jan 2, 2015 at 4:05 PM, Reza Naghibi <[email protected]>
wrote:
Werner, one thing at a time. First, can you please demonstrate how the client
is breaking when using an alias? Aliases, or parents, are used all over the
DDR. Every single device has a parent up until you reach the root device. This
has been the case since the first release of OpenDDR in 2011.
From: Werner Keil <[email protected]>
To: [email protected]; Reza Naghibi <[email protected]>
Sent: Friday, January 2, 2015 9:05 AM
Subject: Re: Improper "aliasing" of devices
Reza,
That ticket did not tell you to "break device-data" which that incomplete entry
did.Everyone in the project (except you maybe;-/) agrees and understands the
value of the BuilderDataSource and DeviceDataSource including the importance of
classes in that file.
You and everybody in the team are obliged to preserve consistency of the 1.x
data branch with W3C DDR specs where all these files and their relations are
defined by.The "duplicate" entry which caused the flaw in Classifier only under
Java 8 was not a data problem, as there can be multiple <value> entries in
BuilderDataSource e.g. <device id="HTC_Touch_Pro2"> <list>
<value>HTC</value> <value>touch_pro2</value>
</list> </device>
And it's neither wrong nor illegitimate to have the same <value> in more than
one builder entry. Making assumptions like a 1:1 relationship are short-sighted
handling by th parser.If you can't handle or understand BuilderDataSource,
please leave it alone and stick to DeviceDataSource, but there one must stick
to the standard, too.
In case you didn't notice, BuilderDataSource also contains Regular Expressions
for some builders, not just simple device Ids. These are used for more precise
detection by the builders, and if a variation in UA was as simple as "SGP312"
vs. "SGP311" that's how W3C DDR deals with such cases, not by breaking the
DeviceData file;-)
<device id="DROID X2" parentId="genericMotorola"> <property
name="model" value="MB870"/> <property name="marketing_name"
value="DROID X2"/> <property name="displayWidth" value="540"/>
<property name="displayHeight" value="960"/> <property
name="mobile_browser" value="Android Webkit"/> <property
name="inputDevices" value="touchscreen"/> <property name="device_os"
value="Android"/> <property name="device_os_version" value="2.2"/>
<property name="dual_orientation" value="true"/> <property
name="ajax_support_javascript" value="true"/> <property
name="ajax_support_getelementbyid" value="true"/> <property
name="ajax_support_inner_html" value="true"/> <property
name="ajax_manipulate_dom" value="true"/> <property
name="ajax_manipulate_css" value="true"/> <property
name="ajax_support_events" value="true"/> <property
name="ajax_support_event_listener" value="true"/> <property
name="image_inlining" value="true"/> <property name="from"
value="oddr"/> </device>
and
<device id="MB870" parentId="genericMotorola"> <property
name="model" value="MB870"/> <property name="marketing_name"
value="DROID X2"/> <property name="displayWidth" value="540"/>
<property name="displayHeight" value="960"/> <property
name="mobile_browser" value="Android Webkit"/> <property
name="mobile_browser_version" value="4.0"/> <property
name="device_os" value="Android"/> <property
name="device_os_version" value="2.2.2"/> <property
name="inputDevices" value="touchscreen"/> <property
name="dual_orientation" value="true"/> <property
name="ajax_support_javascript" value="true"/> <property
name="ajax_support_getelementbyid" value="true"/> <property
name="ajax_support_inner_html" value="true"/> <property
name="ajax_manipulate_dom" value="true"/> <property
name="ajax_manipulate_css" value="true"/> <property
name="ajax_support_events" value="true"/> <property
name="ajax_support_event_listener" value="true"/> <property
name="image_inlining" value="true"/> <property name="from"
value="oddr"/> </device>
As well as Thousands of other device signatures in DeviceData show how this
works, and trying to "shortcut" it with a crippled single line entry didn't.
So if you want to do this properly, please create a proper entry for SGP312
like it's done for all the other device signatures or start with "your own
2.x" structure without discussing requirements and do it there.
Cheers,Werner
Werner Keil | JCP Executive Committee Member, JSR 363 Co Spec Lead | Eclipse
UOMo Lead, Babel Language Champion | Apache Committer | Advisory Board Member,
DWX '15 Twitter @wernerkeil | @UnitAPI | @JSR354 | @AgoravaProj | @DeviceMap |
#EclipseUOMo | #DevOps Skype werner.keil | Google+ gplus.to/wernerkeil
On Fri, Jan 2, 2015 at 7:22 AM, Reza Naghibi <[email protected]>
wrote:
Werner, please undo this commit immediately:
http://svn.apache.org/viewvc/devicemap/trunk/data/device-data/src/main/resources/devicedata/BuilderDataSource.xml?view=diff&r1=1648972&r2=1648973&pathrev=1648973
You just undid this ticket:
https://issues.apache.org/jira/browse/DMAP-86
That ticket was a request from an important user, Infomaker Scandinavia AB.
Secondly, I do not see the connection between that change and your claim that
"aliasing" in device data is breaking the client. I am shocked that you are
being so persistent with very little idea of what is going on. Quite literally,
you are grasping at straws, only now you are committing code. This is a serious
problem.
I closed DMAP-119 and DMAP-120 because the unit tests fail when you link in
device data 1.0.2. The unit tests were designed around 1.0.1. The above
enhancement, DMAP-86, added new devices to 1.0.2. You then proceeded to test
those devices against unit tests written against 1.0.1. So obviously they are
going to fail. This has nothing to do with aliasing.
Werner, this is your final warning. Please undo these changes immediately.
Everything you have done and said tonite leads me to believe you are extremely
confused regarding how the data and API work and what the goal of this project
is. After you revert the changes you made to the data and API, you are not to
commit anymore changes to trunk. Am I clear?
No more arguments Werner. This is very serious. Please just do as I say and
give it a break for a few days.
From: Werner Keil <[email protected]>
To: [email protected]
Sent: Friday, January 2, 2015 12:48 AM
Subject: Improper "aliasing" of devices
Reza/all,
As of data 1.x the attempt of "aliasing" device data by trying to override
a parent device with just a single incomplete line is not working and
illegitimate.
It breaks both clients.
If such "simplified" device data modeling was desired, it has to be done
for 2.x, as it breaks compatibility with the 1.x and W3C data model.
After elliminating the wrong/incomplete line for "SGP312" remaining SGP311
is properly recognized.
Based on Data 1.0.1 it seems the 2 new/wrong entries are ignored, in fact
if below issue
Tests run: 1176, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 1.67 sec
<<< FAILURE! - in org.apache.devicemap.DeviceMapClientTest
testDeviceMapClient[1174](org.apache.devicemap.DeviceMapClientTest) Time
elapsed: 0 sec <<< FAILURE!
org.junit.ComparisonFailure: classification failed for 'Mozilla/5.0 (Linux;
Android 4.3; SGP311 Build/10.4.B.0.577) AppleWebKit/537.36 (KHTML, like
Gecko) Chrome/38.0.2125.102 Safari/537.36' *expected:<[genericAndroid]> but
was:<[SGP311]>*
at org.junit.Assert.assertEquals(Assert.java:115)
at
org.apache.devicemap.DeviceMapClientTest.testDeviceMapClient(DeviceMapClientTest.java:74)
is addressed using data 1.0.2-SNAPSHOT with the correct test data file for
SGP311, the extra line SGP312 in the test file is safely ignored.
Please if needed add a complete record,
This fixes both https://issues.apache.org/jira/browse/DMAP-120 and
https://issues.apache.org/jira/browse/DMAP-121 you bluntly and prematurely
brushed away.
Thanks,
Werner
