TEterhechte·Aug 20, 2014
Here's a good explanation of what the software tries to achieve, from one of the files in the Github repo. Much more informative than the Wired article:
Here's a concrete example: A police department in a city hosts an Excel spreadsheet on their web server called Crime-2013.xls. It contains all of the reported crime so far this year and gets updated every night at midnight with all of the new crimes that were reported each day.
Say you wanted to write a web application that showed all of the crime on a map. To download the new data every night you'd have to write a custom program that downloads the .xls file every night at midnight and imports it into your application's MySQL database.
To get the fresh data imported you can simply delete your entire local crime database and re-import all rows from the new .xls file, a process known as a '
Read the rest
kill and fill'.
But the kill and fill method isn't very robust, for a variety of messy reasons. For instance, what if you cleaned up some of the rows in the crime data in your local database after importing it last time? Your edits would get lost.
Another option is a manual merge, where you try and import each and every row of the incoming Excel file one at a time. If the data in the row already exists in the database, skip it. If the row already exists but the incoming data is a new version, overwrite that row. If the row doesn't exist yet, make a whole new row in the database.
The manual merge can be tricky to implement. In your import script you will have to write the logic for how to check if an incoming row already exists in your database. Does the Excel file have its own Crime IDs that you can use to look up existing records, or are you searching for the existing record by other method? Do you assume that the incoming row should completely overwrite the existing row, or do you try to do a full row merge?
At this point the import script is probably a few dozen lines and is very specific to both the police department's data as well as your application's database. If you decide to switch from MySQL to PostgreSQL in the future you will have to revisit this script and re-write major parts of it.
If you have to do things like clean up formatting errors in the Police data, re-project geographic coordinates, or change the data in other ways there is no straightforward way to share those changes publicly. The best case scenario is that you put your import script on GitHub and name it something like 'City-Police-Crime-MySQL-Import' so that other developers that want to consume the crime data in your city won't have to go through all the work that you just went through.
Sadly, this workflow is the state of the art. Open data tools are at a level comparable to source code management before version control.
https://github.com/maxogden/dat/blob/master/docs/what-is-dat...