From: "Nicolas Frey" <n.frey@proxmox.com>
To: "Dominik Csapak" <d.csapak@proxmox.com>, <yew-devel@lists.proxmox.com>
Subject: Re: [PATCH proxmox-yew-widget-toolkit 3/5] widget: grid picker: keep selected records hidden by the filter
Date: Thu, 05 Nov 2026 11:17:08 +0100 [thread overview]
Message-ID: <DMN7784OLOHL.2IMUYNFOO84ZH@proxmox.com> (raw)
In-Reply-To: <53d706ba-5154-4369-8d67-ccb009adb23a@proxmox.com>
On Mon Nov 2, 2026 at 3:12 PM CET, Dominik Csapak wrote:
> comments inline
>
> On 10/28/26 4:15 PM, Nicolas Frey wrote:
>> the data table drops selected records that do not match the store
>> filter. in a grid picker, the filter is only a search aid, so typing a
>> query deselected all records that did not match it.
>>
>> add an opt-in flag to only drop records that are no longer in the
>> store, using a new function `DataStore::contains_key`, and enable it in
>> the grid picker.
>>
>> Signed-off-by: Nicolas Frey <n.frey@proxmox.com>
>> ---
>> src/state/data_store.rs | 3 ++
>> src/state/store.rs | 4 +++
>> src/state/tree_store/mod.rs | 4 +++
>> src/widget/data_table/data_table.rs | 51 ++++++++++++++++++++++++-----
>> src/widget/grid_picker.rs | 2 ++
>> 5 files changed, 56 insertions(+), 8 deletions(-)
>>
>> diff --git a/src/state/data_store.rs b/src/state/data_store.rs
>> index 90460c2..16b59ea 100644
>> --- a/src/state/data_store.rs
>> +++ b/src/state/data_store.rs
>> @@ -64,6 +64,9 @@ pub trait DataStore: Clone + PartialEq {
>> fn filtered_record_pos(&self, key: &Key) -> Option<usize>;
>> fn filtered_data_len(&self) -> usize;
>>
>> + /// Returns whether the store contains a record with this key, regardless of the filter.
>> + fn contains_key(&self, key: &Key) -> bool;
>> +
>
> This is a bit too narrow for my taste for such an important part of
> the interface. I'd probably like more 'unfiltered' methods
> that mirror the 'filtered' ones. with that, your contains_key
> is just a 'unfiltered_record_pos(_).is_some()'
>
> Also even when adding only the 'contains' method, i'd add the
> 'unfiltered' part, otherwise this can be misunderstood without
> looking at the docs.
>
makes sense, will adjust in a v2, thanks!
>> #[allow(clippy::type_complexity)]
>> fn filtered_data<'a>(
>> &'a self,
>> diff --git a/src/state/store.rs b/src/state/store.rs
>> index 1afe0a4..89a1ad4 100644
>> --- a/src/state/store.rs
>> +++ b/src/state/store.rs
>> @@ -429,6 +429,10 @@ impl<T: Clone + PartialEq + 'static> DataStore for Store<T> {
>> self.filtered_data_len()
>> }
>>
>> + fn contains_key(&self, key: &Key) -> bool {
>> + self.read().record_pos(key).is_some()
>> + }
>> +
>> fn filtered_data<'a>(
>> &'a self,
>> ) -> Box<dyn Iterator<Item = (usize, Box<dyn DataNode<T> + 'a>)> + 'a> {
>> diff --git a/src/state/tree_store/mod.rs b/src/state/tree_store/mod.rs
>> index 5e84b8f..67f0d76 100644
>> --- a/src/state/tree_store/mod.rs
>> +++ b/src/state/tree_store/mod.rs
>> @@ -288,6 +288,10 @@ impl<T: Clone + PartialEq + 'static> DataStore for TreeStore<T> {
>> self.filtered_data_len()
>> }
>>
>> + fn contains_key(&self, key: &Key) -> bool {
>> + self.read().lookup_node(key).is_some()
>> + }
>> +
>> fn filtered_data<'a>(
>> &'a self,
>> ) -> Box<dyn Iterator<Item = (usize, Box<dyn DataNode<T> + 'a>)> + 'a> {
>> diff --git a/src/widget/data_table/data_table.rs b/src/widget/data_table/data_table.rs
>> index 90f37ea..2ee8ed2 100644
>> --- a/src/widget/data_table/data_table.rs
>> +++ b/src/widget/data_table/data_table.rs
>> @@ -237,6 +237,14 @@ pub struct DataTable<S: DataStore> {
>> #[prop_or_default]
>> pub multiselect_mode: MultiSelectMode,
>>
>> + /// In multiselect mode, keep selected records that are hidden by the store filter.
>> + ///
>> + /// By default, the selection is reduced to the records that match the filter, so that
>> + /// actions on the selection only affect visible rows. Enable this if the filter is only a
>> + /// search aid, and must not change the selection.
>> + #[prop_or_default]
>> + pub keep_filtered_selection: bool,
>> +
>
>
> With this i asked myself: 'what is a filtered selection' ?
>
> I think something like 'keep_selection_when_filtered' would be more
> descriptive (even though it's longer).
>
ack
>
>> /// Table scroll callback
>> #[prop_or_default]
>> pub on_table_scroll: Option<Callback<Event>>,
>> @@ -395,6 +403,17 @@ impl<S: DataStore> DataTable<S> {
>> self.allow_deselect = allow_deselect.into_prop_value();
>> }
>>
>> + /// Builder style method to set the keep_filtered_selection flag.
>> + pub fn keep_filtered_selection(mut self, keep: impl IntoPropValue<bool>) -> Self {
>> + self.set_keep_filtered_selection(keep);
>> + self
>> + }
>> +
>> + /// Method to set the keep_filtered_selection flag.
>> + pub fn set_keep_filtered_selection(&mut self, keep: impl IntoPropValue<bool>) {
>> + self.keep_filtered_selection = keep.into_prop_value();
>> + }
>> +
>
> these could be automatically implemented by using the #[builder] macro
>
ah yeah forgot, as all the other ones are written out without the macro
I just went with this. I'll update the others with #[builder] too in a
separate commit
>> /// Builder style method to set the show_header flag.
>> pub fn show_header(mut self, show_header: impl IntoPropValue<bool>) -> Self {
>> self.set_show_header(show_header);
>> @@ -765,6 +784,9 @@ impl<S: DataStore> PwtDataTable<S> {
>> }
>>
>> let mut keys: HashSet<Key> = HashSet::new();
>> + if props.keep_filtered_selection {
>> + keys.extend(selection.selected_keys());
>> + }
>> for pos in 0..record_count {
>> if let Some(key) = props.store.lookup_filtered_record_key(pos) {
>> keys.insert(key);
>> @@ -797,7 +819,11 @@ impl<S: DataStore> PwtDataTable<S> {
>> return;
>> }
>>
>> - let selection_len = selection.len();
>> + // only count visible rows, the selection may contain records hidden by the filter
>> + let selection_len = (0..record_count)
>> + .filter_map(|pos| props.store.lookup_filtered_record_key(pos))
>> + .filter(|key| selection.contains(key))
>> + .count();
>
> since the selection is basically a hashset, we could think
> of exposing a the usual set operations too (intersect, difference, etc)
> and offer a 'filtered_get_keys' that returns a hashset of the existing
> (visible) keys
>
> nothing wrong with this code here, per se but exposing more of the
> selection might benefit us in the future too...
>
ack
>> if record_count == selection_len {
>> self.selection_status = RowSelectionStatus::All;
>> } else if selection_len > 0 {
>> @@ -812,13 +838,22 @@ impl<S: DataStore> PwtDataTable<S> {
>> if let Some(selection) = &props.selection {
>> let mut selection = selection.write();
>> if selection.is_multiselect() {
>> - let mut keys: HashSet<Key> = HashSet::new();
>> - for (_pos, node) in props.store.filtered_data() {
>> - let key = props.store.extract_key(&node.record());
>> - if selection.contains(&key) {
>> - keys.insert(key);
>> - }
>> - }
>> + let keys = if props.keep_filtered_selection {
>> + selection
>> + .selected_keys()
>> + .into_iter()
>> + .filter(|key| props.store.contains_key(key))
>> + .collect()
>> + } else {
>> + props
>> + .store
>> + .filtered_data()
>> + .filter_map(|(_pos, node)| {
>> + let key = props.store.extract_key(&node.record());
>> + selection.contains(&key).then_some(key)
>> + })
>> + .collect()
>> + };
>> selection.bulk_select(keys);
>> } else if let Some(key) = selection.selected_key() {
>> if props.store.filtered_record_pos(&key).is_none() {
>> diff --git a/src/widget/grid_picker.rs b/src/widget/grid_picker.rs
>> index b40ac92..d3ac5ae 100644
>> --- a/src/widget/grid_picker.rs
>> +++ b/src/widget/grid_picker.rs
>> @@ -185,6 +185,8 @@ impl<S: DataStore + 'static> Component for PwtGridPicker<S> {
>> .clone()
>> .key(Key::from("picker-table"))
>> .autoselect(false)
>> + // the filter is only a search aid and must not change the selection
>> + .keep_filtered_selection(true)
>> .cell_configuration(CellConfiguration::new().class("pwt-datatable-cell pwt-pointer"))
>> .hover(true)
>> .header_focusable(false)
next prev parent reply other threads:[~2026-11-05 10:17 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-10-28 15:15 [PATCH proxmox-yew-widget-toolkit 0/5] add multi-select component Nicolas Frey
2026-10-28 15:15 ` [PATCH proxmox-yew-widget-toolkit 1/5] widget: data table: simplify lookup of keys near the cursor Nicolas Frey
2026-11-02 14:12 ` Dominik Csapak
2026-11-06 9:35 ` applied: " Dietmar Maurer
2026-10-28 15:15 ` [PATCH proxmox-yew-widget-toolkit 2/5] widget: data table: only extend explicit select all to new rows Nicolas Frey
2026-11-02 14:12 ` Dominik Csapak
2026-11-05 10:17 ` Nicolas Frey
2026-10-28 15:15 ` [PATCH proxmox-yew-widget-toolkit 3/5] widget: grid picker: keep selected records hidden by the filter Nicolas Frey
2026-11-02 14:12 ` Dominik Csapak
2026-11-05 10:17 ` Nicolas Frey [this message]
2026-10-28 15:15 ` [PATCH proxmox-yew-widget-toolkit 4/5] widget: form: implement multi select field Nicolas Frey
2026-11-02 14:12 ` Dominik Csapak
2026-11-05 10:17 ` Nicolas Frey
2026-11-02 14:23 ` Dominik Csapak
2026-11-05 10:17 ` Nicolas Frey
2026-11-05 10:36 ` Dominik Csapak
2026-10-28 15:15 ` [PATCH proxmox-yew-widget-toolkit 5/5] widget: form: combobox: update doc to point to `MultiSelect` Nicolas Frey
2026-11-02 14:12 ` Dominik Csapak
2026-11-05 10:17 ` Nicolas Frey
2026-11-02 14:12 ` [PATCH proxmox-yew-widget-toolkit 0/5] add multi-select component Dominik Csapak
2026-11-06 7:19 ` Dietmar Maurer
2026-11-06 8:41 ` Dietmar Maurer
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=DMN7784OLOHL.2IMUYNFOO84ZH@proxmox.com \
--to=n.frey@proxmox.com \
--cc=d.csapak@proxmox.com \
--cc=yew-devel@lists.proxmox.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox