Month: November 2017

Please provide the text you want to translate.

Today, I did something I’m proud of (it always happens when you dabble a bit in something that’s not quite your field, and it works out).

In my project, there are currently about a thousand content pages, hundreds of functional components, many of which are used on many pages, with either unique or common configurations. The client updates both the components’ configurations and their composition on the pages, and as often happens with such volumes, fixing something in one place breaks something in another. And the task is to keep all this under control.

What I did: On my server, a graphic snapshot is created for all pages of the site – a set of images with a complete copy of the page at the moment the snapshot is taken. Each picture is stored in folders corresponding to the language, country, site, and hierarchical categories. All are available on the web. Graphic copies of individual page components are stored in a similar manner. Every day. There is a mechanism to visualize changes. Of course, when changes appear where they were not expected, it is both good and bad. Bad, because it means someone did something wrong, and good because it was detected early.

I can’t show anything; it’s a commercial project.

Technologies: python, selenium, chromedriver, imagemagick.

10 km, 5:33, to Wisła, along Wisła, from Wisła

https://www.strava.com/activities/1265475805/shareable_images/map_based?hl=en-US&v=1510078350

Having tea

Georgian restaurant in the Jewish quarter in Poland

An interesting task has arisen. In the Hybris admin panel, there is no way to record all administrator actions so that later, if something happens, we can understand who messed up. Some objects keep their change history and there is a reference to the data changer, but various operations leave at best entries in the log, but never with a reference to a specific administrator.

Analyzing server logs is also futile because admin panels are now built on Single Page Application.

I’ve come up with an idea of how to implement such logging for any system, not just Hybris. I am trying to find such a ready-made tool, but have not succeeded yet. It’s relatively easy to make. If you understand Chrome extensions, we could try to make a PoC together.

The idea is as follows:

1. The administrator logs into the site’s admin panel (not necessarily Hybris).

2. A Chrome extension understands that the opened site is from the list for operation logging and switches to work mode.

3. At certain events, like “data input in a form,” “clicking on an element,” the system performs the operation logging and sends the results to the server. Namely:

a) takes a screenshot of the screen and highlights the area where the element is located (or the entire screen if it’s easier, for starters).

b) adds additional metadata to the screenshot, if able to extract it (for example, for a textarea – the text entered)

c) adds the operator ID.

d) puts the screenshot with metadata in a queue for sending.

e) gradually processes the queue by sending the screenshots to the server

On the server, the screenshots fall into a FIFO stack, accessible for analysis by a super admin if a problem arises.

I see the following disadvantages:

1. Unclear how to implement drag-n-drop, for example. Also unclear how to determine if a user has filled out a form (just entered something and left the textarea—should a screenshot be sent in this case? idk)

2. Not always easy to enrich screenshots with metadata. For example, an onClick handler might be on an empty div while another div may contain the active element with data.

3. A lot of attention to system security

4. Possible system slowdown due to constant screenshot uploads. If actively clicking in the admin panel, outgoing traffic could be about 1Mb/sec. Potentially could cause slowdowns. But still, this is not a very big problem.

If there are experts in Chrome Extensions – welcome. Maybe, we can sketch a PoC together?

P.S. In the comments, they ask why not analyze the requests. Here’s why:

From the bus window today.

2.5 hours by bus, then another 40 minutes by a different one, then 10 km on foot, and I’m in the Tatra Mountains, Morskie Oko. It is said to be one of the most beautiful mountain lakes in the world

Continuing the restaurant theme, I’m dining at my favorite Corleone

I would kill those who make such timezone selections. It’s the twenty-first century, Tesla, gravitational waves, SpaceX has relaunched a spent rocket, a computer has finally beaten humans at Go, yet in Windows for setting a timezone you need to remember the UTC offset or a typical city for it.

Why not incorporate a complete list of major cities into the system and create a convenient directory? There are only about 500 cities in the world with a population over one million, and around 5000 with more than 100,000 inhabitants.