補發 onsite 15 Mar 2018
感覺是備胎了二星期，後來也跪了。。。。
1 Sharpness Value
2 Max island size
Follow up: what if we can flip some cell, return the current max number of island size after the flip…
Union find ba… check leetcode, number of islands series
3. Web Crawler
补充内容 (2018-9-14 22:41):
https://paper.dropbox.com/doc/In … 7npLjRBjrAKA5AOBsaM
其他準備或答案在此，不用找我了
也歡迎在文件中留言 或更正。
當時掛了太傷心，好像忘了補貼這個，那個是Detailed Prep &Solutions：
The same question but change the way it asks… into water level, land elevation.
But, my interview seems not very familiar with the solution, I need to explain him the dp. and the logics all these things…
the i, j , all these indexing make it hard to understand as well…
I make one logic bug and find out during testing… it is I switch the two nested loop
I made wrong…
for(int i=0; i<m…)
for(int j=1; j<n; …)
This is wrong, the outer loop should be the col loop as we move from left to right…
Talk a bit about my proposal on optimised space usage… just store one prev column and use one var called prev…
There is no ask on what if the input is very large etc…
Problem in details:
Let say, I have a water level x;
in the matrix, left to right, find a path that is not get flooded by the water…(means the height/elevation that cell represents should be > x)///
well, in the path, we pick the smallest val, and that smallest val has to be >x…
Then we find the path which give that best value… so we can assue there is always a path if x is not greater or equal to that value…
(same problem as sharpness val then)
No follow up… the space optimised is suggested by me…
Lunch
Demo
Just leetcode question,
I got one small bug regarding visited marking,… I didn’'t mark the starting point as visited… find it out during testing and corrected.
Follow up:
//no need to write exact working code…
//explained Union find … and
// assumed api for find set… and set set … etc…
just write a piece of logic…
First, single thread version…
Talked a bit about DFS, BFS, stack overflow… endless depth etc problem…
Implemented simple BFS, using result set to do visisted checking as well
Multithreading:
Which part is bottle neck: the getURL part… network latency etc… parser etc
I mentioned Master / slave model, and how it can be achieved using ThreadPool, Callable, and check futures…
and metioned it is even more efficient if we don’'t do the syn of these queue, set checking logic just let master manage all these… he seems like this…
Then, he asked, in some language/machine, there is no support of threadPool, how would we do it …
He gave an API… to let you avoid writing some setup code///
setThread(THREAD_COUNT, method)…
implement the method… actually it is the BFS method… but with right locking
There is some back and forth about where to put lock, and
also asked the terminating conditions… (queue is empty and no working thread)
public void crawl() {
while(true){ //this simplify my code for locking / unlock
lock.lock();
if(q.isEmpty() && workingThread<1) {
lock.unlock();
break;
}
//need to use while to guard it
while(q.isEmpty()) {//queue is empty but cannot terminate yet
wait(); //wait will release the lock...
}
String currURL = q.poll();
lock.unlock();
//you don''t want to lock the part that takes most time and that part is not accessing common data
String newURLGathered = getURL(currURL);
lock.lock();
for(each){
add to q & add to result set
}
notifyAll(); //in case there is someone waiting due to empty queue.
lock.unlock();
}
}
他假設有這樣的一個API，你只需要implement 那method，
setThread(THREAD_COUNT, method)…
求solution wang2579@umn.edu
sent you
多谢楼主分享 求crawler的solution 下周也要面这家 crawler没写过有点不稳 yuzhoul dot eecs @ gmail.com
發你了，
楼主求一发crawler的solution谢谢！~
sent you
sent you
vet65hao@126.com 处于兴趣求一份答案 感谢
求solution. 多谢 raloxifene@outlook.com
求链接access approval
Could you send/share web crawler multithreaded solution to mymail.rai@gmail.com?
-Thanks
Hi, Could you share with me the crawler (multithreaded) code at mymail.rai@gmail.com
-Thanks
求solution 谢谢 nydkim@gmail.com