item.get is needed before item.claims can be accessed (on the next line).

We could

1. replace item.claims with a different API call that only gets the list of properties used on the item
2. extend option 1 to be a generic approach to lazy load item data
3. move "has claims for all properties" check further down in the process.

The problem with option 3 is that immediately after this check, harvest_template does a page.get() , and page.get() is probably more expensive than item.get(), at least on English Wikipedia where article text size exceeds typical wikidata item JSON size. This may not be true for smaller wikis where the average article text size is smaller (but I would expect it is true for most of the top 10 wikipedia)

In T76391#800221, @jayvdb wrote:

page.get() is probably more expensive than item.get()

The bot now uses preloading generator, so page.get() is not expensive at all and even not necessary.

Change 370664 had a related patch set uploaded (by Matěj Suchánek; owner: Matěj Suchánek):
[pywikibot/core@master] [IMPR] Share claimit.py logic with harvest_template.py

https://gerrit.wikimedia.org/r/370664

Change 370664 merged by jenkins-bot:
[pywikibot/core@master] [IMPR] Share claimit.py logic with harvest_template.py

https://gerrit.wikimedia.org/r/370664