Angenommen, du arbeitest mit einem Coding-Agenten deiner Wahl (Claude Code, Codex, name your favorite). Während du die anstehende Aufgabe beschreibst, stellst du fest, dass es an anderen Stellen einige relevante Informationen gibt, die der Agent zuerst überprüfen sollte, um einen besseren Kontext für die Erstellung eines Implementierungsplans zu haben.
Wenn du eine komplexe Vorbereitungsaufgabe hast – wie die Suche in Confluence, das Ausführen von CLI-Skripten oder die Recherche im Internet –, führt die Ausführung all dieser Aufgaben in der Hauptsession von Claude Code dazu, dass das Kontextfenster mit HTML-Responses, unübersichtlichen CLI-Ausgaben und irrelevanten Suchergebnissen zugemüllt wird.
Hier ist ein Vorschlag, wie du diese „Context Engineering”-Pipeline mithilfe von Subagenten praktisch steuern kannst, sodass der Hauptagent nur den destillierten, sauberen Kontext erhält.
Du musst die Subagenten nicht manuell im Detail verwalten. Sie agieren als Orchestrator und weisen Ihren Hauptagenten an, spezialisierte Worker zu erstellen.
Du würdest Claude Code eine Eingabeaufforderung wie diese geben:
„Ich möchte [Feature X] in dem Repository erstellen, in dem du dich gerade befindest. Diese Feature soll folgende Funktion erfüllen [Feature Beschreibung].
Bevor du Code schreiben kannst, brauchen wir einen sauberen Implementierungsplan für den du noch Informationen aus verschiedenen Quellen zusammentragen musst. WICHTIG: Verunreinige nicht diese Hauptsession mit diesen Recherchen, sondern erzeuge stattdessen isolierte Subagenten, um den Kontext zu sammeln und nur die relevanten Informationen in diese Hauptsession zurückzugeben:
Folgende Datenquellen wollen wir einbeziehen
- Agent 1 (Confluence): Erstelle einen Agenten, der in Confluence die Dokumentation zur „Auth Architecture API” sucht und die Endpunktspezifikationen zusammenfasst.
- Agent 2 (CLI/Repo): Erstelle einen Agenten, der
grep-und AST-Tools auf dem lokalen Repo ausführt, um alle aktuellen Implementierungen desAuthModulezu finden.- Agent 3 (Web Research): Erstelle einen Agenten, der die aktuellen Best Practices für „OAuth2-Token-Rotation in Next.js” recherchiert.
- Agent 3 (Repository Analysis): Erstelle einen Agenten, der das aktuelle Repository nach ähnlichen Implementierungen durchsucht und sich eine Übersicht verschafft, wo welche Teile der Implementierung platziert und welche existierenden Artefakte (Spezifikation, Code, Tests, etc.) angepasst werden müssen.
- Halte in jedem Fall die Design Principles und Constraints aus deiner ARCHITECTURE.md ein.
Warte, bis alle drei fertig sind. Nimm dann ihre zusammengefassten Zusammenfassungen, synthetisiere sie und schreiben sie eine saubere
IMPLEMENTATION_PLAN.md. Zeige mir nur die endgültige Markdown-Datei.”
Im Hintergrund verwendet Claude Code das Tool sessions_spawn.
Die übergeordnete Sitzung (Ihr Hauptfenster von Claude Code) erhält drei saubere, prägnante Textblöcke. Sie sieht niemals die unübersichtlichen Grep-Fehler oder den 10-seitigen Confluence-Sidebar-HTML-Code. Sie verwendet diese sauberen Zusammenfassungen, um den Implementierungsplan zu entwerfen.
Hast du gute Erfahrungen gesammelt wie du den Context deines Orchestrators sauber halten kannst? Teil sie gern mit mir auf Linkedin oder als Github Issue.
Im nächsten Post werde ich über Erfahrungen bei der Qualitäts-Sicherung durch Skills und Subagents eingehen.
]]>I’ve been thinking about this topic since my article on Innovation Tokens [1] in 2016. In past articles I often used the term Small Experiments. I connect a lot of semantics to that term which is all about enabling innovation in larger organisations which might otherwise struggle with Innovator’s Dilemma. I want to share with you how Small Experiments can be a catalyst for Innovation in any organization.
Some might call them “Proof of concept” (POC), yet, I favor experiment for the original meaning of the term:
“A test under controlled conditions that is made to demonstrate a known truth, examine the validity of a hypothesis, or determine the efficacy of something previously untried.”
- The American Heritage® Dictionary of the English Language, 5th Edition.
The term „Experiment“ is mostly familiar in the context of scientific research. It means to set up a hypothesis, an assumption and validate whether it’s true or not. The result of an experiment is exactly that - a learning whether a question can be answered TRUE under the tested conditions.
Now, how does such certainty help me, you might ask. I very often read about the VUCA world in which it is very tough to take strategic decisions that have long-term impact due to the uncertainty it bares. Small Experiments address exactly that uncertainty by saying “Okay, what conditions or circumstances are in scope for that experiments. In other words: Which probabilities must be considered so that we have sufficient reliability and can take that decision with certainty?”
And that doesn’t only speed up business decisions but for any risky decision. We validate assumptions and hypothesis by trying them out in practice until we have enough certainty to take the decision. This is an alternative to spending countless hours in meetings just discussing the matter.
To be clear, I’m not saying “STOP DISCUSSIONS!”. It’s about having a discussion to agree on which question, hypothesis, or assumption we must have an answer to. The more precise the question, the smaller we can draw the scope of our Experiment. A small Scope is important for multiple reasons:
We probably all experienced discussions with larger groups of people. One comment can spark a new aspect, provoke another comment, a new aspect… and before you notice, we’re discussing a different topic.
With a defined small scope to consider, we can reduce the number of expert people that have to involved to answer the experiment’s question. Having a small team is known to have a lot of positive side effects such as a smaller number of people is easier to coordinate, builds trust faster, agrees quicker, etc. Of course, the neurodiversity of a small group can become an issue. If the participants neglect a perspective that is relevant for the context of the experiment. Yet, that’s something we can already address when building the group.
The scope of an Experiment should be small enough, that everyone in the company can immediately understand
Showing potential impact of an experiment is important because you address the sorrows and concerns of employees who are afraid that change e.g. might cost their jobs. Surely that’s not the case for every single experiment, yet, being transparent about where you’re heading brings clarity and builds trust and again provides you with the chance to get feedback on concerns you’d never truly known existed in people’s minds. Comprehensibility helps to spark such thought processes in people and thus increases the chance for valuable feedback. This feedback might contain relevant variables which haven’t been on the radar of the experiment so far.
The other positive impact of accessibility and comprehensibility are also important because it shortens and simplifies the time it takes to explain the Experiment to people outside the experiment group. This is helpful e.g. when the experiment group realizes they need additional support from a person outside their group.
That’s why I call them Small Experiments: The risk is small, the potential benefit is high and that’s how I address what I’d call my “life’s learning” from my colleague, Stefan Tilkov. Stefan taught me to always look at the positive potential of an idea and how we could enable that instead of drowning myself in what might all go wrong. If we’re exploring an unknown path to the top of the mountain, it’s only normal we’ll find that “ah, we cannot take this way. Let’s go back and head to the left”. Small experiments support the same Management of Uncertainties for organizations that want to take yet unexplored paths to the top. Yes, we need to be able to mitigate risks and failures because this enables us to make a move at all. Fear of failure otherwise brings us to a standstill.
There are different steps we could take to move towards establishing small experiments. And certainly there are still things to be looked after e.g. where the limits of scientific methods are and where methods of complexity management can support them. Thanks for taking the time to read this.
[1] https://blog.heusingfeld.de/organisation/innovation-tokens-jaxenter/
PS: I’d like to thank my colleagues who helped me reflect my thinking on the above over the course of time and finally help me put these thoughts into a series of articles: Tim Riemer, Phillipp Tepel, Jessica Paas, Phillip Ghadir, Eberhard Wolf, Julius Ganns, and Stefan Tilkov.
]]>In case you also come into this great situation "I upgraded my Ubuntu to 16.04 LTS and suddenly my owncloud installation is a mess", don’t panic. I’m about to share my mess with you so you can pick the perils.
It all started with Ubuntu 14.04LTS an innocent do-release-upgrade … which didn’t want to be aborted - twice. So it convinced to "yes, ok, let’s do it now…". After the normal upgrade process and a reboot my system was back and working fine - I thought, until I noticed mails from my monitoring system, that the owncloud webUI wasn’t available.
A quick check confirmed "yes, the webUI is not available". And even after the hard part of fixing the nginx configuration - yes, I’m running owncloud behind nginx - and manually installing php7.0-fpm - did I mention Ubuntu 16.04 replaced php5 with php7 EXCEPT FOR THE FPM PART?! -, all websites worked, but Owncloud wouldn’t!
Taking a very close look into nginx error.log and I suddenly noticed that the index.php was missing from /var/www/owncloud! In fact, it seems during the Ubuntu Upgrade process for an unreproducible reason all .php files in /var/www/owncloud got deleted! But hey, no worries, let’s apt-get install --reinstall! I double checked that my sources list was correctly pointing to owncloud stable for Ubuntu 16.04 (see https://download.owncloud.org/download/repositories/stable/owncloud/) and apt-get installed owncloud 9.1.3 on my machine without mentioning any problems.
Unfortunately my monitoring system was still unhappy. "Ah, of course, the 'manual upgrade' process", I thought, and ran sudo -u www-data /var/www/owncloud/occ upgrade only to get this wonderful output:
ownCloud or one of the apps require upgrade - only a limited number of commands are available
You may use your browser or the occ upgrade command to do the upgrade
Set log level to debug
Exception: Updates between multiple major versions and downgrades are unsupported.
Update failed
Maintenance mode is kept active
Reset log levelWhen I opened the webUI, it showed the same error message. But additionally it included a link to the immensely helpful https://central.owncloud.org/t/updates-between-multiple-major-versions-are-unsupported/815 which pointed out that I had indeed jumped from Owncloud 8 directly to Owncloud 10 - I skipped the 9.0 version!
Ok, let’s cut this short: It occurred to me that the owncloud team has no (documented) support for such a situation - at least I don’t consider https://doc.owncloud.org/server/9.0/admin_manual/maintenance/package_upgrade.html#upgrading-across-skipped-releases helpful at all! I assume your next step was also to downgrade owncloud to the next minor version of your current release and by trying this, you jumped into the next fun situation: the owncloud-server package has a dependency on PHP5! IF SOMEONE CAN TELL ME WHY THIS IS CONSIDERED REASONABLE, I’M REALLY CURIOUS?
Anyway after reading through http://askubuntu.com/questions/762221/ubuntu-16-04-upgrade-killed-owncloud-and-many-other-things-but-right-now-i-n I finally came to this post http://askubuntu.com/questions/767158/ubuntu-16-04-cannot-install-owncloud-server which has a very interesting hint: The owncloud-files package has almost no depencies, is thus easy to downgrade and includes the updater! WOOHOO, JACKPOT!
So, I started by modifying my /etc/apt/sources.list.d/owncloud.list to Ubuntu 8.2
sh -c "echo 'deb http://download.owncloud.org/download/repositories/8.2/Ubuntu_16.10/ /' > /etc/apt/sources.list.d/owncloud.list"
and then forcing the downgrade and step-by-step upgrade by running
$ sudo apt-get update && sudo apt-get install owncloud-files=8.2.9-1.1
(...)
$ sudo -u www-data /var/www/owncloud/occ upgrade
(...)
$ sh -c "echo 'deb http://download.owncloud.org/download/repositories/9.0/Ubuntu_16.10/ /' > /etc/apt/sources.list.d/owncloud.list"
$ sudo apt-get update && sudo apt-get upgrade
(...)
$ sudo -u www-data /var/www/owncloud/occ upgrade
(...)
$ sh -c "echo 'deb http://download.owncloud.org/download/repositories/stable/Ubuntu_16.10/ /' > /etc/apt/sources.list.d/owncloud.list"
$ sudo apt-get update && sudo apt-get upgrade
(...)
$ sudo -u www-data /var/www/owncloud/occ upgrade
(...)It took a while but worked pretty well. So in the end I’d like to thank John Orion for sharing his idea about owncloud-files and the owncloud team for providing this option - even though you don’t mention it on the pages where people are seeking for help. I hope this is helpful for others, struggling with upgrades.
Now I’m on to fix the remaining issues: Currently I’m facing a situation where all requests with index.php are returned with a HTTP302 redirect without index.php - which of course fails (e.g. /cloud/index.php/login is redirected to /cloud/login). NOTE: I’m running owncloud behind nginx with a contextroot other than /owncloud - as if I was asking for trouble, right? I’ll update this blog, in case I stumble upon something.
If you find anything to comment, optimize or fix about this, please leave me a github issue. Always think "If I have this issue, someone else might sooner or later have it, too" - and you can help that someone. Thanks a lot!
For those who want to follow along, I opened a Github issue under https://github.com/owncloud/core/issues/26869 and provided a pull-request for the remaining necessary changes here https://github.com/owncloud/core/pull/26882
The article I wrote on "Innovation Tokens: Hilfsmittel bei Informatikerromantik und Technologie-Überflutung" which was published in the German "Business Technology" magazine, has just been republished on jaxenter.de! You can find it at https://jaxenter.de/innovation-tokens-50334 in German language.
]]>I wrote a post on the innoQ blog about "Per request debugging with Log4j 2 filters".
The bare minimum that should be available are the application’s log messages, so that’s probably the simplest thing to start with. From what I saw, in most production environments log output is unfortunately reduced to the absolute minimum for the sake of system performance and resource efficiency. In those environments it would be ok to increase the logging for a certain portion of the traffic, e.g. just for one incoming HTTP request, but this seems to be quite complicated. In this post I’d like to show you that it isn’t.
In the Java ecosystem we are free to choose from a variety of logging frameworks to help us get the job done. There is the time-honoured Apache Log4J, it’s so-called successor Logback and the SLF4J API which became quite popular. But with this post I’d like to direct your attention to Log4J’s new major version release Apache Log4j 2. Apart from greater performance and numerous bugfixes, Log4j 2 also comes with some new and optimised features which we can use to accomplish our task.
The first thing we need is a place to intercept the HTTP request, extract the appropriate HTTP header and add it to Log4j’s ThreadContext. The ThreadContext is a "Mapped Diagnostic Context" (MDC) which is also known from other logging frameworks. Below you will find a naive implementation of a ServletFilter which does exactly this, assuming that the parameter to enable debugging is a boolean HTTP header like x-debug-enabled: true. Of course you can use any other request parameter you like.
package com.innoq.blog.samples;
import org.apache.logging.log4j.ThreadContext;
import javax.servlet.*;
import javax.servlet.http.HttpServletRequest;
import java.io.IOException;
/**
* ServletFilter to extract request data and add it to Log4j's ThreadContext.
*/
public class RequestDebugFilter implements Filter {
public static final String DEBUG_HEADER = "x-debug-enabled";
@Override
public void doFilter(ServletRequest request, ServletResponse response, FilterChain chain) throws IOException, ServletException {
try {
// NOTE: Parsing for a "Boolean" makes sure that ThreadContext does not contain the random value that someone put into the HTTP header!
Boolean debugHeaderEnabled = Boolean.valueOf(((HttpServletRequest)request).getHeader(DEBUG_HEADER));
if (debugHeaderEnabled) {
ThreadContext.put(DEBUG_HEADER, debugHeaderEnabled.toString());
}
chain.doFilter(request, response);
} finally {
ThreadContext.remove(DEBUG_HEADER);
}
}
/* ... further methods abbreviated... */
}Once we got the parameter in our ThreadContext, we don’t need any further code changes, but can solely focus on the Log4j 2 configuration. In our Log4j configuration we can now make use of the DynamicThresholdFilter to generally change the log level to TRACE in case the parameter "x-debug-enabled" is set to "true". A simple yet complete log4j2.xml file doing this could look like the following:
<?xml version="1.0" encoding="UTF-8"?>
<Configuration monitorInterval="30">
<!-- This filter is used to generally enable TRACE logging if "x-debug-enabled" in the ThreadContext has the value "true" -->
<DynamicThresholdFilter key="x-debug-enabled" onMatch="ACCEPT" onMismatch="NEUTRAL">
<KeyValuePair key="true" value="TRACE"/>
</DynamicThresholdFilter>
<Appenders>
<Console name="CONSOLE" target="SYSTEM_OUT" >
<JSONLayout compact="true" charset="utf-8" eventEol="true" properties="true"/>
</Console>
</Appenders>
<Loggers>
<Root level="info">
<AppenderRef ref="CONSOLE"/>
</Root>
</Loggers>
</Configuration>With this in hand we might even decide that we want to write those special HTTP requests to a different log file. This can be done by adding the following snippet to our log4j2.xml:
<RollingFile name="DEBUGLOGFILE" fileName="logs/debug.log" filePattern="logs/debug-%d{yyyyMMdd_HH}-%i.log.gz">
<Filters>
<ThreadContextMapFilter onMatch="ACCEPT" onMismatch="DENY">
<KeyValuePair key="x-debug-enabled" value="true"/>
</ThreadContextMapFilter>
</Filters>
<Policies>
<OnStartupTriggeringPolicy />
<TimeBasedTriggeringPolicy />
</Policies>
<PatternLayout charset="UTF-8">
<PatternLayout pattern="[[%d{ISO8601}||%X||%-5p||%t||%c{1.}||%m]]%n"/>
</PatternLayout>
</RollingFile>In this case we are using a ThreadContextMapFilter to ACCEPT log messages only in case the x-debug-enabled parameter in the ThreadContext map is set to true. Otherwise the log message is DENIED meaning not written to this RollingFile appender.
We’ve just shown how to enable TRACE logging with Log4j 2 on a per-request basis, but of course this isn’t limited to HTTP requests. Any parameter in the ThreadContext can be used as a trigger. Therefore it is possible to debug batch jobs which have a certain flag, WebSocket connections or any other thread-based processing.
The Log4j 2 manual has a lot more information and also explains use cases for new features. I recommend you read through it if you want to gain a deeper understanding. If you happen to read German, you might also be interested in the introductory article on Log4j 2 "Logging konsolidieren und Performance gewinnen" by my colleagues Stefan and Phillip.
I am running a self-hosted owncloud instance for more than a year now. Back in August I decided to setup a new Ubuntu 14.04 server at home and tried to run owncloud with a PostgreSQL backend - which turned out to be a bad decision.
Quite often I noticed the following error in my owncloud.log file:
{"app":"index","message":"Doctrine\\DBAL\\DBALException: An exception occurred while executing 'SELECT \"fileid\", \"storage\", \"path\", \"parent\", \"name\", \"mimetype\", \"mimepart\", \"size\", \"mtime\",\n\t\t\t\t\t \"storage_mtime\", \"encrypted\", \"unencrypted_size\", \"etag\", \"permissions\"\n\t\t\t\tFROM \"oc_filecache\" WHERE \"storage\" = ? AND \"path_hash\" = ?':\n\nSQLSTATE[25P02]: In failed sql transaction: 7 ERROR: current transaction is aborted, commands ignored until end of transaction block","level":4,"time":"2014-11-26T15:28:16+00:00"}
{"app":"PHP","message":"Cannot modify header information - headers already sent at \/var\/www\/owncloud\/lib\/private\/response.php#83","level":3,"time":"2014-11-26T15:28:16+00:00"}Later on I found out that it occurred when owncloud tried to scan and index the files in the data folder. The way I got to know this is, I noticed that some of the files I uploaded were not displayed in owncloud but were present on the filesystem. So when I ran
sudo -u www-data /usr/bin/php5 /var/www/owncloud/occ files:scan ahe -vthe command also ended in this error message:
[Doctrine\DBAL\DBALException]
An exception occurred while executing 'SELECT "fileid", "storage", "path", "parent", "name", "mimetype", "mimepart", "size", "mtime",
"storage_mtime", "encrypted", "unencrypted_size", "etag", "permissions"
FROM "oc_filecache" WHERE "storage" = ? AND "path_hash" = ?':
SQLSTATE[25P02]: In failed sql transaction: 7 ERROR: current transaction is aborted, commands ignored until end of transaction block
Exception trace:
() at /var/www/owncloud/3rdparty/doctrine/dbal/lib/Doctrine/DBAL/DBALException.php:47
Doctrine\DBAL\DBALException::driverExceptionDuringQuery() at /var/www/owncloud/3rdparty/doctrine/dbal/lib/Doctrine/DBAL/Statement.php:140
Doctrine\DBAL\Statement->execute() at /var/www/owncloud/lib/private/db/statementwrapper.php:63
OC_DB_StatementWrapper->execute() at /var/www/owncloud/lib/private/db.php:235
OC_DB::executeAudited() at /var/www/owncloud/lib/private/files/cache/cache.php:128
OC\Files\Cache\Cache->get() at /var/www/owncloud/lib/private/files/cache/homecache.php:62
OC\Files\Cache\HomeCache->get() at /var/www/owncloud/lib/private/files/cache/cache.php:367
OC\Files\Cache\Cache->remove() at /var/www/owncloud/lib/private/files/cache/scanner.php:163
OC\Files\Cache\Scanner->removeFromCache() at /var/www/owncloud/lib/private/files/cache/scanner.php:272
OC\Files\Cache\Scanner->scanChildren() at /var/www/owncloud/lib/private/files/cache/scanner.php:284
OC\Files\Cache\Scanner->scanChildren() at /var/www/owncloud/lib/private/files/cache/scanner.php:284
OC\Files\Cache\Scanner->scanChildren() at /var/www/owncloud/lib/private/files/cache/scanner.php:284
OC\Files\Cache\Scanner->scanChildren() at /var/www/owncloud/lib/private/files/cache/scanner.php:207
OC\Files\Cache\Scanner->scan() at /var/www/owncloud/lib/private/files/utils/scanner.php:125
OC\Files\Utils\Scanner->scan() at /var/www/owncloud/apps/files/command/scan.php:57
OCA\Files\Command\Scan->scanFiles() at /var/www/owncloud/apps/files/command/scan.php:81
OCA\Files\Command\Scan->execute() at /var/www/owncloud/3rdparty/symfony/console/Symfony/Component/Console/Command/Command.php:244
Symfony\Component\Console\Command\Command->run() at /var/www/owncloud/3rdparty/symfony/console/Symfony/Component/Console/Application.php:897
Symfony\Component\Console\Application->doRunCommand() at /var/www/owncloud/3rdparty/symfony/console/Symfony/Component/Console/Application.php:191
Symfony\Component\Console\Application->doRun() at /var/www/owncloud/3rdparty/symfony/console/Symfony/Component/Console/Application.php:121
Symfony\Component\Console\Application->run() at /var/www/owncloud/console.php:43
require_once() at /var/www/owncloud/occ:11
[PDOException]
SQLSTATE[25P02]: In failed sql transaction: 7 ERROR: current transaction is aborted, commands ignored until end of transaction block
Exception trace:
() at /var/www/owncloud/3rdparty/doctrine/dbal/lib/Doctrine/DBAL/Statement.php:138
PDOStatement->execute() at /var/www/owncloud/3rdparty/doctrine/dbal/lib/Doctrine/DBAL/Statement.php:138
Doctrine\DBAL\Statement->execute() at /var/www/owncloud/lib/private/db/statementwrapper.php:63
OC_DB_StatementWrapper->execute() at /var/www/owncloud/lib/private/db.php:235
OC_DB::executeAudited() at /var/www/owncloud/lib/private/files/cache/cache.php:128
OC\Files\Cache\Cache->get() at /var/www/owncloud/lib/private/files/cache/homecache.php:62
OC\Files\Cache\HomeCache->get() at /var/www/owncloud/lib/private/files/cache/cache.php:367
OC\Files\Cache\Cache->remove() at /var/www/owncloud/lib/private/files/cache/scanner.php:163
OC\Files\Cache\Scanner->removeFromCache() at /var/www/owncloud/lib/private/files/cache/scanner.php:272
OC\Files\Cache\Scanner->scanChildren() at /var/www/owncloud/lib/private/files/cache/scanner.php:284
OC\Files\Cache\Scanner->scanChildren() at /var/www/owncloud/lib/private/files/cache/scanner.php:284
OC\Files\Cache\Scanner->scanChildren() at /var/www/owncloud/lib/private/files/cache/scanner.php:284
OC\Files\Cache\Scanner->scanChildren() at /var/www/owncloud/lib/private/files/cache/scanner.php:207
OC\Files\Cache\Scanner->scan() at /var/www/owncloud/lib/private/files/utils/scanner.php:125
OC\Files\Utils\Scanner->scan() at /var/www/owncloud/apps/files/command/scan.php:57
OCA\Files\Command\Scan->scanFiles() at /var/www/owncloud/apps/files/command/scan.php:81
OCA\Files\Command\Scan->execute() at /var/www/owncloud/3rdparty/symfony/console/Symfony/Component/Console/Command/Command.php:244
Symfony\Component\Console\Command\Command->run() at /var/www/owncloud/3rdparty/symfony/console/Symfony/Component/Console/Application.php:897
Symfony\Component\Console\Application->doRunCommand() at /var/www/owncloud/3rdparty/symfony/console/Symfony/Component/Console/Application.php:191
Symfony\Component\Console\Application->doRun() at /var/www/owncloud/3rdparty/symfony/console/Symfony/Component/Console/Application.php:121
Symfony\Component\Console\Application->run() at /var/www/owncloud/console.php:43
require_once() at /var/www/owncloud/occ:11
files:scan [--all] [user_id1] ... [user_idN]As this is a PostgreSQL error, I searched the owncloud forums for other people’s experience with owncloud + PostgreSQL. Unfortunately it seems that owncloud doesn’t play too nicely with PostgreSQL. :(
Looking back I had an owncloud setup with MySQL before and cannot remember having had any issues with it. So I decided to take the easy way and find out how to migrate my owncloud data from PostgreSQL to MySQL. This was when I stumbled upon this great little command in the owncloud console.php: db:convert-type
This little gem did the trick for me so that after executing the following command, my data had automatically been migrated over to MySQL and the owncloud configuration changed accordingly. As it’s been so nice to me I thought, I leave it here for others coming up with the same question:
sudo -u www-data /usr/bin/php5 /var/www/owncloud/console.php db:convert-type --all-apps mysql owncloud localhost owncloudFor those of you interested in the impact of a company’s culture, I wrote an article about the company culture at innoQ on their blog. The article is on the company website and I’d really appreciate your feedback e.g. on writing style and content.
]]>Lately I noticed a growing dissatisfaction with my daily workflow. And this was not only because my last blog post has been almost a year ago. No, it’s related to a noticable drop in productivity. I figured this was due to my curiosity dragging me onto way more interesting topics than either my time frame or my toolset could master. Unfortunately trying to keep up-to-date on those topics created loads of distractions which almost resulted in an inability to focus on a single task for a longer amount of time
Lucky me, I’m surrounded by a bunch of awesome people whom I could talk to about this and my pending decisions.
First of all I’ll try to get over that "writer’s block" and write small posts like this one about things on my mind.
Second of all I will have to reduce the quantity of my output on Twitter and skip some of my projects to achieve a level of quality that is finally satisfactory for me personally.
Regarding the toolset I decided to give DayOne another try. It’s a journal/ diary app that supports MarkDown (yes, I’d favor Asciidoc), syncs via iCloud, reminds me of stuff and has a few other nice features. The only downsides are it’s Apple-only and it doesn’t support WebDAV-sync which would allow automatic publishing to my blog.
That’s it for my plans. My next post will hopefully be quite soon as I’m currently preparing my JavaOne talk (CON4952) with my dear colleague, Martin Eigenbrodt, which touches some interesting aspects of #MicroServices and distributed systems. Let’s see how it will work out.
The Android Team’s switch from the Eclipse based Android Developer Tools to the IntelliJ IDEA based "Android Studio" at Google I/O is pretty well-known. But what isn’t too well-known is that they also changed the default android project directory structure as they switched their build system from ant to gradle.
Further information on the new structure and the new build system are available here: http://tools.android.com/tech-docs/new-build-system/user-guide
I decided to go with the new Gradle-based structure as I thought it was more future proof. I have to admit that I regret this badly! As I’m not too familiar with Gradle, yet, I had a pretty bad time trying to understand the customizations brought in by the gradle-android plugin. Additionally most of the information available on the net is specific to Eclipse/ Ant based projects and therefore slightly different for the new structure.
download Android SDK
extract it to e.g. ~/dev/android-sdk-macosx
modify your .bash_profile and add
export ANDROID_HOME=YOUR_ANDROID_SDK_PATH_HERE
create local.properties
IntelliJ project creation wizard > creates eclipse/ Ant structure
Android Studio project creation wizard
Pivotal Labs' "AndroidIntelliJStarter" https://github.com/pivotal/AndroidIntelliJStarter
These utilities turned out to be buggy or outdated and partly messed up the working configuration. Don’t use them!
What turned out to be really helpful, were Quality-Tools-for-Android and FEST Android by SQUARE. You should really have a look at Square’s repositories at Github as there is a lot of goodness for the new Android world lying around!
1. Before you try things out, go RTFM.
It might take you two weeks to read https://developer.android.com/design/index.html and https://developer.android.com/develop/index.html but don’t dare to start before this! Except if you like to start-over or rewrite your whole Application. ;)
2. Before you use a library, assure that it works with Gradle!
dip = Density-independent Pixels
sp = Scale-independent Pixels
The compiler accepts both "dip" and "dp", though "dp" is more consistent with "sp".
A: Probably your ".idea/gradle.xml" is messed up. See this bug report: https://code.google.com/p/android/issues/detail?id=56750 to fix it
A: The reason for this is that the android.jar contains a version of JUnit prior to 3.8! You can fix this by opening the "Project structure" settings > your module > Dependencies and putting the "Android Platform" dependency all the way down the list.
A: Yes, there is a great post by @tbeernot describing this pain. But to the rescue there is miglayout for android!
In the meantime there is a migration guide and a Tips & Tricks section available on the Android Developer pages!
In the meantime there are a few really awesome videos available on Gradle and Android, that I’d like to share with you
Introduction to Android Studio - Gradleware
Gradle for Android - Hans Dokter, CEO @Gradleware
Learnings from building Twitter for Android - Jonathan Le, Lead Dev Twitter for Android @Twitter
Five Lessons to Electrify your Gradle plugins - Jake Ouellette, Crashlytics Team @Twitter
I have been invited by the DOAG to give a talk on "Enterprise Integration with GlassFish" at their SIG Middleware event.
This event will take place in Duesseldorf on Oct, 8th 2013. It is a meetup of the German Oracle customers using Oracle middleware products. The event will feature a number of great talks around GlassFish 4.0.