Use all bus data

[Katsute]

Feature

Anytime a bus is requested:

1. Check the cache
2. Pull data for all busses
3. Parse
4. Fulfill the request

Reason

Current implementation runs a new request for every bus which is not efficient for enterprise. Requesting all busses then parsing is more efficient when submitting multiple requests.

[Katsute]

Use similar binary search implemented in #90 against bus line not possible

Store raw data in cache, DO NOT STORE OBJECTS; will cause memory issues.

< 10,000 busses

https://bt.mta.info/wiki/Developers/SIRIVehicleMonitoring

Please note that the calls made without either a VehicleRef or LineRef produces quite a load on the system, so use them sparingly. Any developers found to be making repeated calls (e.g. at less than 30 second intervals) for all vehicles in the system may find their key revoked.

[Katsute]

Only this subclass needs to be modified

OneMTA/src/main/java/dev/katsute/onemta/MTAService.java

Line 52 in f3dfe30

final class BusService {

Stop monitoring should derive values from vehicle request if possible.

[Katsute]

JSON parser needs to be massively optimized to avoid memory issues; use JIT (just-in-time) parsing, where we load immediate keys into memory, then parse the subsequent object when requested.

---

This count needs to be optimized to only run once: not possible due to split logic

OneMTA/src/main/java/dev/katsute/onemta/Json.java

Line 129 in f3dfe30

final long count = Regex9.count(quotes.reset(after));

https://sites.google.com/site/gson/streaming

https://stackoverflow.com/a/11876086

Regex may actually be more inefficient since we are rescanning the string multiple times. Use character loop stream instead.

---

Return a method to parse for:

OneMTA/src/main/java/dev/katsute/onemta/Json.java

Lines 197 to 200 in f3dfe30

	else if(arrayMatcher.group("array") != null) // open new array
	list.add(openArray(reader, json));
	else if(arrayMatcher.group("map") != null) // open new map
	list.add(openMap(reader, json));

and

OneMTA/src/main/java/dev/katsute/onemta/Json.java

Lines 235 to 238 in f3dfe30

	else if(mapMatcher.group("array") != null) // open new array
	obj.set(key, openArray(reader, json));
	else if(mapMatcher.group("map") != null) // open new map
	obj.set(key, openMap(reader, json));

[Katsute]

@mashiro-san create a branch upgrade-json

[ghost]

@Katsute I have created a new branch upgrade-json@f3dfe30

[Katsute]

• get methods need to be optimized so synchronized isn't blocking threads
• pull bus stop alert info from gtfs feed
• derive stop vehicles from all vehicles

[Katsute]

Use CopyOnWriteArrayList to allow concurrent reads and locked writes.

ConcurrentMap?

In order to make write thread safe, have synchronized write block also check if its expired before fetching (any current active writes will make the result not expired, any queued writes will see that the resource is no longer expired and will return the new one instead).

Get first then check if expired to prevent concurrency issues.

[Katsute]

@mashiro-san create a new branch upgrade cache

[ghost]

@Katsute I have created a new branch upgrade-cache@ebf9c8e

[Katsute]

Possibly use #100 to implement this.