mirror of
https://github.com/gristlabs/grist-core.git
synced 2024-10-27 20:44:07 +00:00
Comments, adjust for loop
This commit is contained in:
parent
46456fd439
commit
cd8af59079
@ -186,11 +186,15 @@ export class Cursor extends Disposable {
|
||||
// NOTE: _cursorEdited
|
||||
// We primarily update cursorEdited counter from a this._properRowId.subscribe(), since that catches updates
|
||||
// from many sources (setCursorPos, arrowKeys, save/load, filter/sort-changes, etc)
|
||||
// However, when deleting a row with several sections linked together, there can be a case where this fails:
|
||||
// When GridView.deleteRows calls setCursorPos to keep cursor from jumping after delete, the observable
|
||||
// However, there's some cases where we user touches a section and properRowId doesn't change. Obvious one is
|
||||
// clicking in a section on the cell the cursor is already on. This doesn't change the cursor position, but it
|
||||
// SHOULD still update cursors to use that section as most up-to-date (user just clicked on a cell!), so we do
|
||||
// it here. (normally is minor issue, but can matter when a section has rows filtered out so cursors desync)
|
||||
// Also a more subtle case: when deleting a row with several sections linked together, properRowId can fail to
|
||||
// update. When GridView.deleteRows calls setCursorPos to keep cursor from jumping after delete, the observable
|
||||
// doesn't trigger cursorEdited(), because (I think) _properRowId has already been updated that cycle.
|
||||
// This caused a bug when several viewSections were cursor-linked to each other and a row was deleted.
|
||||
// We can explicitly call cursorEdited again here. It'll cause cursorEdited to be called twice sometimes,
|
||||
// This caused a bug when several viewSections were cursor-linked to each other and a row was deleted
|
||||
// NOTE: Calling it explicitly here will cause cursorEdited to be called twice sometimes,
|
||||
// but that shouldn't cause any problems, since we don't care about edit counts, just who was edited latest.
|
||||
this._cursorEdited();
|
||||
|
||||
@ -208,7 +212,7 @@ export class Cursor extends Disposable {
|
||||
}
|
||||
|
||||
// Should be called whenever the cursor is updated
|
||||
// _EXCEPT FOR: when cursor is set by linking
|
||||
// EXCEPT FOR: when cursor is set by linking
|
||||
// this is used to determine which widget/cursor has most recently been touched,
|
||||
// and therefore which one should be used to drive linking if there's a conflict
|
||||
private _cursorEdited(): void {
|
||||
|
@ -156,27 +156,42 @@ function isValidLink(source: LinkNode, target: LinkNode) {
|
||||
return false;
|
||||
}
|
||||
|
||||
// Walk backwards along the chain of ancestors
|
||||
// - if we hit a non-cursor link before reaching target, then that would be an illegal cycle
|
||||
// - when we hit target, we've verified that this is a legal cycle, so break
|
||||
// (ancestors further up the hierarchy past target don't matter, since once we set target.linkSrcSec = src.sec,
|
||||
// they would stop being ancestors of src)
|
||||
// NOTE: we're guaranteed to hit target before the end of the array (because of the `if(...includes...)` above)
|
||||
// ALSO NOTE: isAncestorSameTableCursorLink may be 1 shorter than ancestors, but it's accounted for by the above
|
||||
let i = 0;
|
||||
for (;;) {
|
||||
// We know our ancestors cycle back around to ourselves
|
||||
// - lets walk back along the cyclic portion of the ancestor chain and verify that each link in that chain is
|
||||
// a cursor-link
|
||||
|
||||
// e.g. if the current link graph is:
|
||||
// A->B->TGT->C->D->SRC
|
||||
// (SRC.ancestors):[5][4] [3] [2][1] [0]
|
||||
// We're verifying the new potential link SRC->TGT, which would turn the graph into:
|
||||
// [from SRC] -> TGT -> C -> D -> SRC -> [to TGT]
|
||||
// (Note that A and B will be cut away, since we change TGT's link source)
|
||||
//
|
||||
// We need to make sure that each link going backwards from `TGT -> C -> D -> SRC` is a same-table-cursor-link,
|
||||
// since we disallow cycles with other kinds of links.
|
||||
// isAncestorCursorLink[i] will tell us if the link going into ancestors[i] is a same-table-cursor-link
|
||||
// So before we step from i=0 (SRC) to i=1 (D), we check isAncestorCursorLink[0], which tells us about D->SRC
|
||||
let i;
|
||||
for (i = 0; i < source.ancestors.length; i++) { // Walk backwards through the ancestors
|
||||
|
||||
// Once we hit the target section, we've seen all links that will be part of the cycle, and they've all been valid
|
||||
if (source.ancestors[i] == target.section.getRowId()) {
|
||||
// We made it! All is well
|
||||
break;
|
||||
break; // Success!
|
||||
}
|
||||
// If we've hit the last ancestor and haven't found target, error out (shouldn't happen, we checked for it)
|
||||
if (i == source.ancestors.length-1) { throw Error("Array doesn't include targetSection"); }
|
||||
|
||||
// We're not done with the cycle. Check next link, if it's non-same-table-cursor-link, the cycle is invalid.
|
||||
if (!source.isAncestorSameTableCursorLink[i]) { return false; }
|
||||
|
||||
i++; // Walk back
|
||||
// Check that the link to the preceding section is valid
|
||||
// NOTE! isAncestorSameTableCursorLink could be 1 shorter than ancestors!
|
||||
// (e.g. if the graph looks like A->B->C, there's 3 ancestors but only two links)
|
||||
// (however, if there's already a cycle, they'll be the same length ( [from C]->A->B->C, 3 ancestors & 3 links)
|
||||
// If the link doesn't exist (shouldn't happen?) OR the link is not same-table-cursor, the cycle is invalid
|
||||
if (i >= source.isAncestorSameTableCursorLink.length ||
|
||||
!source.isAncestorSameTableCursorLink[i]) { return false; }
|
||||
}
|
||||
|
||||
// If we've hit the last ancestor and haven't found target, error out (shouldn't happen!, we checked for it)
|
||||
if (i == source.ancestors.length) { throw Error("Array doesn't include targetSection"); }
|
||||
|
||||
|
||||
// Yay, this is a valid cycle of same-table cursor-links
|
||||
}
|
||||
|
||||
|
Loading…
Reference in New Issue
Block a user