With the pandemic everyone's got their own challenges. For me it became particularly hard to switch to online grocery shopping.
In my typical order I'd buy groceries for 7-10 days and it will contain 60-80 items in a cart. When I do my shopping offline I usually don't plan much and go through all the departments and pick up stuff as I go. It's rare when I forget to buy something because it catches my eye, but more often I buy something that I don't need (well, that's how supermarkets designed to work).
Online is a different story. I can't reproduce the same behaviour and go through the whole catalog. So I needed to plan my meals. I'd written down ingredients for my recipes on sticky notes and combined them to make an order.
This was really tedious. After a few orders I noticed that it was kind of repetitive. Although my wife will tell you that I just was lazy and didn't want to do boring stuff, the developer in me said that it's time to automate this and never solve the same problem again.
That's how Cooklang was born.
About Cooklang
I though what if I store my recipes in Markdown-like text files and tag ingredients with @ symbol? Like that:
That will make recipe files human and machine readable.
Given these recipe files computers can create a shopping list and actually do much more: calculate nutrition values, costs or whatever given that information for ingredients provided.
And as it's stays human readable, that means I can be agile and store my recipes in git and perfect them over time (like code refactoring).
Also because it's just a simple text format I avoid "vendor lock-in" problem and still can use my recipes the same way when I retire. My recipes are mine, forever.
About tools
Having only a language specification isn't really helpful. Yes, I can store my recipes on GitHub and own them, but that's it.
I imagined that it would be nice to have many small applications which can understand the language and do their own small thing very well: calculate calories, shopping, costs, smart meal planning, etc. I was really excited.
I read a wonderful book Crafting Interpreters by Robert Nystrom and did a few experiments and I created a simple parser and CLI app. As I want to focus here on how I automated shopping I might do another post about the parser and CLI if anyone interested, so I skip all the details.
So CLI can understand Cooklang files, extract ingredients, and group them:
$ cook shopping-list \
> Neapolitan\ Pizza.cook \
> Root\ Vegetable\ Tray\ Bake.cook
BREADS AND BAKED GOODS
breadcrumbs 150 g
DRIED HERBS AND SPICES
dried oregano 3 tbsp
dried sage 1 tsp
pepper 1 pinch
salt 25 g, 2 pinches
FRUIT AND VEG
beetroots 300 g
carrots 300 g
celeriac 300 g
fresh basil 18 leaves
garlic 3 gloves
lemon 1 item
onion 1 large
red onion 2 items
thyme 2 springs
MEAT AND SEAFOOD
parma ham 3 packs
MILK AND DAIRY
butter 15 g
egg 1 item
mozzarella 3 packs
OILS AND DRESSINGS
Dijon mustard 1 tsp
Marmite 1 tsp
cider 150 ml
olive oil 3 tbsp
OTHER (add new items into aisle.conf)
tipo zero flour 820 g
PACKAGED GOODS, PASTA AND SAUCES
vegetable stock 150 ml
water 530 ml
TINNED GOODS AND BAKING
cannellini beans 400 g
chopped tomato 3 cans
fresh yeast 1.6 g
redcurrant jelly 1 tsp
I also can set output format to json or yaml and feed this into other programs. Or just use good old plain text output manipulation and pipes in shell. It has a few more features like a web-server to explore the recipes with cooking mode and make shopping list in the browser.
I migrated a bunch of recipes from scattered sources into Cooklang format and stored them on GitHub repository https://github.com/dubadub/cookbook.
My grocery shopping approach
Unfortunately, I haven't got far after that and fully automated shopping yet. I still need to do some manual steps because my shop doesn't provide any API access (surprise!).
That's how my process looks now:
1/ I generate a list of all ingredients. I added directories with symlinks to recipes which represent a meal plan. So it's easy to generate a list of ingredients for the whole directory like that:
$ cook shopping-list --only-ingredients ./Plan\ I
2/ I remove from the list whatever I have at home.
3/ I paste all the ingredients to a multi search input on the shop's web-site:
4/ I manually go through each item and add to a cart 😓.
5/ Done!
It has a manual step, but still:
- much faster than before;
- has less cognitive load to my brain;
- more precise.
I'm really happy with this precision part. I noticed that I now cover all my needs but not over consume. Eco-sufficiency in some way.
What's next
I solved only 80% of my problem and I want to do more. I'm thinking to create a mapping between ingredients from my recipes and links to them at my shop's web-site. That will allow me to use curl or Selenium to complete the problem.
Top comments (1)
Alexey, I literally just signed up to a Dev.to account that I will probably never use again, just so I can tell you that I think this idea is a great one. I found this post because I was specifically looking for online grocery stores that already do this, and think it's crazy that they haven't done so yet. Great work buddy.