Add prefix checking before displaying list of matched articles

[cz2h]

Attempt to address the issue described in issue #184 .

Add an early return if either the follow conditions holds:

• The search prefix input box is empty. (Which prevents loading search result after redirecting)
• The dirEntry array is non-empty and is not the result from latest prefix search. (Will handle the empty array in the origin code)

[mossroy]

The implementation looked like a good idea at first sight.
But, second thought, I think it won't be reliable (and may introduce bugs) because comparing the first dirEntry url with the search field is not safe.
The current search implementation uses several "variants" of the search field to do the binary search with (and also first trims it). See function findDirEntriesWithPrefix of ZIMArchive. So we can not simply compare it. Doing a case insensitive comparison might cover most cases, but also could give false-positives.
Another problem is that the search algorithm might change in the future. For example to allow to search the string anywhere in the title, or even for full-text search. We won't be able to duplicate this logic here.

A proper fix for #184 (that could also cover #89 as a refactoring of #587) would be to pass a kind of state through all the callbacks, that could be compared with the expected state.
But it might be complicated to implement.
This state could be an object with 2 properties : the type of current action (searching, or going to an article), and an id for the current action (search prefix for a search, article url for an article). Or maybe this state might even be an integer, incremented on each user action. This needs to be thought about more deeply.

[cz2h]

Thanks for the hints in addressing this issue.

I managed to use the latestUserAsynAction as the state to store user's latest action that might cause false displayed. I compare the state to perform early return in readArticle and searchDirEntriesFromPrefix since these are the two callbacks that may cause false displays.

The only problem I found while testing is after searching and clicking an article, the search result might come back too fast and block the action of going to the clicked article but I am assuming this rarely happens.

[mossroy]

The way to implement that might be good, but it lacks testing and re-reading.
Please take more time to review/test your changes before asking for our reviews, else I'll give up.

In ServiceWorker mode, the main page does not show up, and any attempt to open an article gives a :

isDirEntryExpectedToBeDisplayed is not defined

There also are many typos in the comments.
More minor remarks :

• Regarding the wording, I think you should use the word "async" instead of "asyn". The function isThisCallbackExpectedToBePerformed would have a more meaningful name with something like isTheLatestAction
• in the functions that get the new action as a parameter, I would have put it as the last parameter instead of the first. Because it's not the most important parameter, and might be optional : if no action is passed, we don't do any check (like before)

To help testing, you might introduce delays in all the async actions you try to cover, with something like setTimeout. You could even commit it, to help our testing too. And you would remove these delays before the merging

[mossroy]

@zengchu2 : do you intend to finalize this PR?

[kelson42]

No feedback from the developer. I think this PR will have to be closed or finished by someone else.

[mossroy]

Closing this PR. The branch should be kept because I think the implementation was on the right track anyway