# A few things

email: here

#music
#photo
#programming

Don't read #tech articles except you really want to.

Some of my projects:
BigPicture
Jeux d'orgues
SamplerBox
Ojourdui

## Why I finally won't go for open-source analytics tool (for now)

You discovered Google Analytics a few years ago (a webmaster tool to see how many visits on your websites), and used it efficiently. But, you know, Google-centralized internet, etc. and then you thought "Let's go self-hosted and open-source!". And then you tried Piwik and Open Web Analytics.

I did the same. After a few months, here are my conclusions.

Open Web Analytics has a great look, close to Google Analytics, but every week, I had to deal with new issues:

• first I discovered that a gigantic table was growing in the MySQL database:
| owa                | owa_request                   |       4.44 |
| owa                | owa_click                     |       5.30 |
| owa                | owa_domstream                 |     238.28 |
+--------------------+-------------------------------+------------+ 

Nearly 250 MB analytics data in 2 weeks (for only a few small websites), this means more than 6 GB of analytics data per year in the MySQL database! ... or even 60 GB per year if you have 100k+ pageviews. That's far too much for my server. This was (nearly) solved by disabling Domstream feature. (Ok Domstream is a great feature, but I would have liked to know in advance that this would eat so much in the database).

• then, once this was solved, a few weeks later I saw this big table coming again (then I had to reset the JavaScript tracking codes for every website)

• today I've seen that a new table in the OWA database was very big (747 MB in a few weeks!)

| owaa               | owa_queue_item                |     747.92 |
+--------------------+-------------------------------+------------+   
• some other issues: login impossible from Chrome in certain situations, unique visitors count wrong when using PHP tracker (sometimes, each new visit / refresh of the page is considered as a new visitor), time-range menu not displayed at all (display stuck on 1-week range) in some cases, etc.

I'm not saying OWA is bad: Open Web Analytics is a good open-source solution, but if and only if you have time to spend, on a regular basis, on configuration issues, which I sadly don't have.

I tried Piwik very quickly. It really is a great project but:

• it doesn't offer a direct view of what I was looking for out-of-the box, i.e. clear charts for every website à la Google Analytics (I can't really describe what's the problem, but the user interface isn't handy for me)

• maybe there's an easy fix for this, but the interface is very slow

So, conclusion:

### Analytics, unsolved problem.

I'm still looking for a lightweight self-hosted solution. Until then, I'll probably have to use Google Analytics again.

PS: No offence meant: most of my work is open-source too, and I know that it takes time to build a stable mature tool. This post is just reflecting the end-of-2016 situation.

## Somme d'exponentielles concernant la fonction de Möbius

Au cours de mon Master 2, en 2007, j'ai eu l'occasion de considérer une somme d'exponentielles concernant la fonction de Möbius:

$$S(x, \theta) = \sum_{n \leq x} \mu(n) e^{2 i \pi n \theta}.$$

En suivant Maier et Sankaranarayanan, il s'agissait de comparer plusieurs preuves du résultat suivant.

Théorème. Soit $\theta$ un nombre irrationnel de type $1$. Alors pour tout $\varepsilon > 0$, on a $$S(x,\theta) \ll x^{4/5 + \varepsilon},$$

où le type d'un irrationnel $\theta$ est défini par

$$\eta = \sup \{\delta > 0 : \liminf_{q \rightarrow \infty} q^\delta \| q \theta \| = 0 \}.$$

et $\| x \|$ est la distance d'un réel $x$ au plus proche entier.

Le mémoire Sur une somme d'exponentielles concernant la fonction de Möbius contient la démonstration de ce théorème ainsi qu'un contenu (très) introductif aux caractères de Dirichlet, fonctions $L$.