Except that a mutation must be synchronous whereas an action can contain asynchronous code, I find it difficult to establish some guidelines on how to correctly choose where the logic is supposed to be implemented.
The Vuex doc describes the mutations as transactions, which makes me think that each mutation must ensure as much as possible state coherence, potentially doing multiple changes in the store within the same mutation.
Let’s take for example the following state, storing documents that can either be loading or loaded:
const state = {
loading: []
loaded: []
}
const mutations = {
setLoaded(state, doc) {
state.loaded.push(doc)
// Remove doc from loading:
const index = state.loading.find(loadingDoc => doc.id === loadingDoc.id)
if (index !== -1) {
state.loading.splice(index, 1)
}
}
}
Here the mutation ensures that the doc is either in loading or loaded state, not both. The problem doing so is that the mutations could eventually end up having a lot of logic, which could make them more difficult to debug through the Vuex dev tools.
Some prefer to keep the mutations as simple as possible (eg. only one state change per mutation), but then it’s up to the actions (or the components in case they directly call mutations) to ensure state coherence, which could be more error-prone since, in the above example, you would need to remove the doc from loading state each time you call the “setLoaded” mutation.
As for data validation, unluckily we have no simple way to do it in mutations since mutations cannot return anything. So an action calling a mutation has no way to know that the data is invalid. Eg.:
const mutations = {
setValue(state, intValue) {
if (intValue < 0) {
// error
return
}
state.value = intValue;
}
}
Thus, it seems that data validation needs to be done in actions, with mutations assuming that the data they receive are correct.
What are your thoughts about this, and more generally about actions and mutations? Do you have guidelines on how to properly use them from your own experience?