mlc_ucla

The language archive sites, mlc and ucla, share the same codebase. They share a common Flask Blueprint with project-specific overrides for each site. The sites use SQLite databases with FTS5 tables for full-text search. Note that these are separate from the SQLite database that is a direct copy of the project's original FileMaker Pro database.

Running this site locally

Set up a Python virtual environment. Clone the repo, install modules:

python3 -m venv venv
source venv/bin/activate
git clone https://github.com/uchicago-library/mlc
pip install --upgrade pip
pip install -r requirements.txt

Get a copy of local.py from another developer or from one of the production servers and place it in the mlc directory. Get a copy of either meso.db or ucla.db from one of the production servers and update the path to the file in local.py. Get glottolog_lookup.json, glottolog_language.ttl, tgn.ttl, and the project triples (meso.big.yyyymmdd.ttl or ucla.big.yyyymmdd.ttl) from one of the production servers, and update the locations of those files in local.py too.

Set an environmental variable for the FLASK_APP:

export FLASK_APP=mlc

Or

export FLASK_APP=ucla

Start a development server on localhost:5000:

flask run

Building the Website Database

The website uses an SQLite database for full text search and browse. You'll need a set of linked data triples to build the database- check the production databases for files named "ucla.big.yyyymmdd.ttl" and "meso.big.yyyymmdd.ttl". If these files aren't available see below for instructions on getting data out of FileMaker Pro and into these triples.

Once you have triples in hand, set the FLASK_APP environmental variable and APP variable in local.py to either 'mlc' or 'ucla' to build the appropriate database. Then run the following command:

flask build-db

Note that when you run this command, you'll get a large number of ISO8601Errors- this is because of dirty data in our triples, you can ignore these. The mlc database takes about 10 minutes to run on my dev machine, and the ucla database takes longer.

Translating

• Strings can be labelled in templates with {% trans %}string to be translated{% endtrans %} or in the code with gettext(u'string to be translated') or lazy_gettext(u'string to be translated') if outside a request
• Two identical strings will be labeled together for translation.
• Strings can be injected into all templates through @app.context_processor. There is a dict() that is already being injected.
• The doc for the translation package is Flask Babel
• sheet being used for translation https://docs.google.com/spreadsheets/d/18m-8sN6Gqu6HFgZNIp1QST-3hy_ul_sm3QQCr_z4W6k/edit?usp=sharing
• po2csv is being used to convert the output .po files to .csv for easy translation. That's why the translation-toolkit required package.
• The current strings were translated using =GOOGLETRANSLATE()function from the Google Sheet.

extract all strings pybabel extract -F babel.cfg -k lazy_gettext -o messages.pot .

create the first translation file or update the translation from an updated messages.pot. In case of confusion, one can delete the root folder translations and start over. pybabel init -i messages.pot -d translations -l es pybabel update -i messages.pot -d translations

The .po file can be used to translate strings directly, or it can be converted into CSV and translated elsewhere. convert po file to CSV po2csv .\translations\es\LC_MESSAGES\messages.po .\translations-csv\messages-to-translate.csv

translate CSV file into new file named messages-translated.csv in the same location

convert CSV back to po csv2po .\translations-csv\messages-translated.csv .\translations\es\LC_MESSAGES\messages.po

compile the translations pybabel compile -d translations

Converting FileMaker to Linked Data

Getting data from FileMaker to the triplestore is a multi-step process. Because OCHRE is the new definitive location for this data, I don't expect this process to ever be used again, however it may be useful for future projects involving FileMaker Pro.

To get data out of FileMaker Pro and into linked data triples, start by exporting the FileMaker database as a sequence of XLSX files. Then import those XLSX files into a local SQLite database, and finally produce triples from that SQLite database.

You need to use FileMaker 18 to work with this database. Do not upgrade to FileMaker 19.

Scripts for importing data are in the data_import directory.

Exporting the FileMaker database as a sequence of XLSX files

1. Save the exported database in a local_db directory. I rename each database to include a datestamp, like LA_ACS.yyyymm.fmp12.
2. Create a new export directory using the following format: export.<yyyy>.<mm> where yyyy and mm are the current year and month.
3. Open the database. The encryption password and password for the user "Admin" are the same- after you have entered the encryption password be sure to log in as the "Admin" user.
4. import.py contains a dictionary of filenames/layouts and database tables. For each:
   1. Go to the "Items" layout in FileMaker.
   2. Go to File > Manage > Layouts.
   3. Click "New", "Layout/Report" in the lower left corner.
   4. Under "Show records from" select the basename of the filename from import.py.
   5. Under "Layout Name" enter export_<basename>.
   6. Below, select "Computer" and "Table".
   7. Click "Finish".
   8. Under "Add Fields", select all fields.
   9. Click "OK".
   10. Go to the "Items" layout.
   11. Select the new layout, "export_<basename>".
   12. Go to File > Export Records.
   13. Under "Save As" enter "<basename>.xlsx".
   14. Save the table export in your export.<yyyy>.<mm> directory.
   15. Under "Type" select "Excel Workbooks (.xlsx)".
   16. Be sure the "Use field names as column names in first row" boolean is checked. It's ok to leave all other fields blank. Click "Continue..."
   17. Under "Specify Field Order for Export", click "Move All". Click "Export". Repeat for each file.
5. Run python check_FM_to_XLSX_export.py to check the export directory. This script will check to be sure that files are named correctly, and it will report back on differences between the number of rows and columns in the exported data. It's normal for the number of rows in export data to differ when the data is being edited.

Importing data from XLSX to SQLite

Run import.py to import the XLSX outputs from FileMaker into SQLite tables.

Check the import

Run check_import.py.

Build triples

Run the following commands to export SQLite data as triples: time python sqlite_to_triples.py --mesoamerican | python kernel_metadata.py > mlp.big.yyyymmdd.ttl time python sqlite_to_triples.py --ucla | python kernel_metadata.py > ucla.big.yyyymmdd.ttl (as of February 2023, the ucla export takes longer, and runs in about 5.6 hours.)

find_value_in_database.py

Find which database tables and fields contain a particular value. This command can be handy if you see a value in the FileMaker GUI and you're not exactly sure where it's being stored in the underlying database.

Diagram relationships in the data

Look at diagram_has_part_relationships.py for an example of how to do this.

noid_to_panopto_mapping.py

Create a NOID to Panopto mapping, for the ARK resolver.

tables.py

List tables and fields in the SQLite database. Use -v for verbose output.