This answered a question I didn't even think to ask -- "what if the cells have different ranks?"
Thanks! On Tue, Dec 12, 2017 at 5:39 PM, bill lam <[email protected]> wrote: > In J dictionary, verb > > The shape of a result is the frame (relative to the cells to which the verb > applies) catenated with the shape produced by applying the verb to the > individual cells. Commonly these individual shapes agree, but if not, they > are first brought to a common rank by introducing leading unit axes to any > of lower rank, and are then brought to a common shape by *padding* with an > appropriate *fill* element: space for a character array, 0 for a numeric > array, and a boxed empty list for a boxed array. For example: > > i."0 s=: 2 3 4 >'I';'was';'here' > 0 1 0 0 I > 0 1 2 0 was > 0 1 2 3 here > > > > On Dec 13, 2017 7:59 AM, "TongKe Xue" <[email protected]> wrote: > >> Both explanations are great. Unfortunately, I asked an XY problem. Let >> me re-ask as follows: >> >> I am trying to write a mini interpreter for J. >> >> How do we execute "i. (2 2 $ 1 2 3 4)" ? >> >> After parsing, and evaluating the (), we are left with: >> >> 1. execute monad "i." on tensor of (shape: [2 2], data: [[1 2] [3 4]]) >> >> 2. Next, we look up the rank of "i." and see that it is [1 _ _] >> >> 3. Therefore, we split [[1 2] [3 4]] into a frame, containing two cells: >> cell1 = [1 2] >> cell2 = [3 4] >> >> 4. We execute i. on each cell, getting: >> cell1-output = [0 1] >> cell2-output = [[0 1 2 3] [4 5 6 7] [8 9 10 11]] >> >> 5. At this point, I am expecting: error! tensor shape mismatch >> >> 6. Instead, J appears to just 0 pad the cells until they are of the same >> size. >> >> Is there some fundamental principle why the 0 padding is happening, or >> does J have a hard coded piece of logic that says: >> when merging the output-cells of a frame, if the output-cells are of >> different shape, 0-pad them all until they are of the same shape ? >> >> Thanks, >> --TongKe >> >> >> >> On Tue, Dec 12, 2017 at 4:35 PM, Brian Schott <[email protected]> >> wrote: >> > The rank of monadic i. is 1 as determined by the following. >> > i. b. 0 >> > 1 _ _ >> > >> > Consider the following where the rows and column lengths of i. 1 2 are >> > made to match those of i. 3 4. >> > >> > (i. 1 2),:i. 3 4 >> > 0 1 0 0 >> > 0 0 0 0 >> > 0 0 0 0 >> > >> > 0 1 2 3 >> > 4 5 6 7 >> > 8 9 10 11 >> > >> > But that may not be explanation enough. >> > >> > On Tue, Dec 12, 2017 at 5:49 PM, TongKe Xue <[email protected]> wrote: >> > >> >> Hi, >> >> >> >> >> >> I understand what (2 2 $ 1 2 3 4) does. >> >> I understand what i. 1 2 does >> >> I understand what i. 3 4 does. >> >> >> >> I have read http://www.jsoftware.com/help/ >> jforc/loopless_code_i_verbs_ >> >> have_r.htm#_Toc191734331 >> >> >> >> I understand the concept of verb-rank, of frames + cells, of >> >> "promoting one frame to another if they share the same prefix." >> >> >> >> I don't understand how the 0 padding in >> >> >> >> i. (2 2 $ 1 2 3 4) works >> >> >> >> >> >> What is the mechanism by which 0-padding is happening? >> >> >> >> >> >> Thanks, >> >> --TongKe >> >> >> >> >> >> ==== >> >> >> >> 2 2 $ 1 2 3 4 >> >> >> >> 1 2 >> >> >> >> 3 4 >> >> >> >> i. 1 2 >> >> >> >> 0 1 >> >> >> >> i. 3 4 >> >> >> >> 0 1 2 3 >> >> >> >> 4 5 6 7 >> >> >> >> 8 9 10 11 >> >> >> >> i. (2 2 $ 1 2 3 4) >> >> >> >> 0 1 0 0 >> >> >> >> 0 0 0 0 >> >> >> >> 0 0 0 0 >> >> >> >> >> >> 0 1 2 3 >> >> >> >> 4 5 6 7 >> >> >> >> 8 9 10 11 >> >> ---------------------------------------------------------------------- >> >> For information about J forums see http://www.jsoftware.com/forums.htm >> > >> > >> > >> > >> > -- >> > (B=) <-----my sig >> > Brian Schott >> > ---------------------------------------------------------------------- >> > For information about J forums see http://www.jsoftware.com/forums.htm >> ---------------------------------------------------------------------- >> For information about J forums see http://www.jsoftware.com/forums.htm > ---------------------------------------------------------------------- > For information about J forums see http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
