Kera likes line segments. He has nn line segments, and each line segment can be described by $l_i$ and $r_i$ on the number axis. Moreover, each line segment Kera has its preferred value $val_i$.

Then there are $m$ operations:

• $(op = 1, l_i, r_i, val_i)$ — Add a line segment $(l_i,r_i,val_i)$ on the number axis, according to the above.
• $(op = 2,L,R)$ — Query the maximum difference of preferred values of all line segments that are fully contained by the interval represented by $[L, R]$.

In addition, a line segment $(l_i, r_i, val_i)$ fully contained by the interval $(L,R)$ means: $L \leq l_i \leq r_i \leq R$.

The first line contains two integer $n$ and $m$.

Each of the next nn lines contains $3$ integers $l_i , r_i$ and $val_i$.

Each of the next $m$ lines contains $4$ integers $(op = 1, a_i, b_i, val_i)$ or $3$ integers $(op = 2,A,B)$.

Specifically, in order to get the real input, you need to create a variable named $\text{LastAns}$ with an initial value of $0$.

For each operation $(op = 1)$, $l_i = a_i xor \text{LastAns}, r_i = b_i xor \text{LastAns}$.

For each operation $(op = 2)$, $L = A xor \text{LastAns}, R = B xor \text{LastAns}$, then $\text{LastAns}$ updated to result of this query.

In addition, xor means binary exclusive OR.

$1 \leq n, m \leq 100000, 1\leq l_i \leq r_i \leq 3000, 0 \leq val_i \leq 1e9, 1\leq L \leq R \leq 3000$.

For each operation $(op = 2)$, print an integer to represent the maximum difference of preferred values.

It is guaranteed that each query fully contains at least one line segment.

The real sample input:

3 5
2 3 1
5 6 4
4 10 5
2 2 5
2 2 6
2 1 10
1 6 7 20
2 1 10