[GUI Plugin] Action Chains

[un_pogaz]

Originally Posted by ownedbycats

Some of the actions have conditionals (for example, run Extract ISBN only if book has no existing ISBN identifier). It doesn't quite work right when some book matches conditionals and others don't.

Yeah, I enconter the same problem. It's really confusing to see that the condition depends exclusively on the exact book wich on you open the Action Chains context menu (or the last/first selected) to check the execution conditions, instead of acting as a filter.
If this book matches the condition, the action will be performed on all selected books, even those that do not match and on which you did not want to perform this action.

And I don't think a real filter feature would be possible, because Action Chains often uses the selected books to retrive the id to perform their own action, so it real not depending of Action Chains that the Calibre actions are all or nothing. A real filter feature would involve deselecting and selecting dynamically wich rows is needed before run the Calibre actions, then redo that at each step of the chain.
... I think it's possible (Calibre is powerfull like that), but it would be terribly complicated, so I will give it no hope.

[un_pogaz]

Okay. By checking the code, I found a possible way.

`gui.current_view().get_selected_ids()` allow to retrive the book id of the selected row, and `gui.current_view().select_rows()` allow to select the book base on their id `class calibre.gui2.library.views.BooksView. Additionaly, using this two function allow to mimic and preserve the selection order.

Beside @capink, get_selected_ids() it probably a better ways to retrive the book id than you current way.
Damn it, if you had GitHub or similiar, I could work on proposing a patch, to either transform the condition on a true filter, or add a new filter feature.
I'll see what I can do, but it won't be as clean to share you the full files like that.

[capink]

Originally Posted by ownedbycats

Some of the actions have conditionals (for example, run Extract ISBN only if book has no existing ISBN identifier). It doesn't quite work right when some book matches conditionals and others don't.

Originally Posted by un_pogaz

I could work on proposing a patch, to either transform the condition on a true filter, or add a new filter feature.
I'll see what I can do, but it won't be as clean to share you the full files like that.

This issue and the solution for it has been addressed before. As you can see in the linked post, you can easily modify the selection (into a subset of the already selected books) based on whatever criteria you want. Conditions were intended as a quick way to determine whether a chain/action will run based on simple criteria that does not take much time to run, because chain conditions are executed every time the Action Chains menu is dynamically created. Any more additional criteria can be done as illustrated in the linked post.

The solution in the post even allows for saving the initial selection into a variable, so that you can restore them mid-chain if you want, and allow for more flexible control.

[un_pogaz]

Interesing, but I think that a formal implementation of a filter system would be a big plus, certainly with its own column/settings.
Also, if I understand your idea of wanting to dynamically display the menu entry based on simple condition criteria and so limit this to a single metadata object for performance reason, I think that inside the chains themselves, this is too confusive. It's certainly too late to replace this column with a proper filter setting, but in any case, I really encourage the addition of such thing.

Else, I worked of the idea and so here my proposal:

Spoiler:

in class chains.Chain._run_loop()

somewhere at the start

# store selected books at the start of the chains
self.chain_selected_ids = self.gui.current_view().get_selected_ids()
self.previous_selected_ids = self.chain_selected_ids

after

do_run = self.check_action_conditions(chain_link.get('condition_settings', {}))

append

# filtre the book for running action
filter_settings = chain_link.get('filter_settings', {})
if not filter_settings.get('template'):
    # No conditions set, use all selected ids
    action_selected_ids = self.chain_selected_ids
else:
    action_selected_ids = []
    for book_id in self.chain_selected_ids:
        mi = self.db.new_api.get_proxy_metadata(book_id)
        if self.check_conditions(filter_settings, mi=mi):
            action_selected_ids.append(book_id)

# action_selected_ids can only contain a subset of chain_selected_ids
# don't perform a new select_rows() if the previous seletion was the same than chain_selected_ids
if len(action_selected_ids) != len(self.chain_selected_ids) or len(self.previous_selected_ids) != len(self.chain_selected_ids):
    # change_current=True because the previous current_row can probaly not match the conditions
    # and so the validate() function can fail and abort the action
    self.gui.current_view().select_rows(action_selected_ids, change_current=True, scroll=False)
self.previous_selected_ids = action_selected_ids

and to insert in the finaly block

# full reset seleted books
if len(self.previous_selected_ids) != len(self.chain_selected_ids):
    self.gui.current_view().select_rows(self.chain_selected_ids, change_current=True, scroll=False)

[chaley]

Originally Posted by un_pogaz

Else, I worked of the idea and so here my proposal:
[/SPOILER]

FWIW: I'm currently implementing these changes to the base formatter.

• ids = selected_books()
• name = selected_cell(). Returns the lookup name.
• A 'with' statement that changes the book the template is using for the statement's block. The same local variables are available to the inner and outer block.

Example:

program: 
	ids = selected_books();
	col = selected_column();
	res = '';
	for id in ids:
		with id:
			# prints information from the book with identifier 'id'
			print($title, field(col));
			res = (if res then res & ',' fi) & field(col);
			template('program: print("inside template", $title)')
		htiw
	rof;
	list_remove_duplicates(res, ',')

Internal template functions and stored templates called from inside the 'with' block use the book with id 'id'.

EDIT: I'm also going to implement "show_in_dialog()" that renders an html string.

[un_pogaz]

Originally Posted by chaley

FWIW: I'm currently implementing these changes to the base formatter.

• ids = selected_books()
• name = selected_cell(). Returns the lookup name.
• A 'with' statement that changes the book the template is using for the statement's block. The same local variables are available to the inner and outer block.

EDIT: I'm also going to implement "show_in_dialog()" that renders an html string.

Oh, it would be inded quite usefull to test all the selected books inside the singular Template call of the condition settings. But that dosn't solve the issue to perform some actions only on a specific subset on seleted books.

Else, if retrive the seleted book ids is good, I think it would be much more useful to have a function 'search_books()' that use the search syntax.

ids = search_books('series:"='&field('series')&'"')

rlst = ''
for id in ids:
	with id:
		rlst = rlst & character('newline') & $title
	htiw;
rof;
show_in_dialog(rlst) # show all the books in a series

And let's even be bold, adding a 'all_ids_books()', that way, we can run the entire library.
(this two function would also greatly benefit to have a optional argument virtual libray)

Damn, it will make the GPM incredible powerfull, able to execute an return big data for a one of execution. Python template are great, but GPM is much simpler and enough for 90% of case, and the 'with' statement will make it a "complet" language in a some way.
But now that I think about it, I don't think the template editor is designed for that kind use, since it executes the code at each edit... it would become a hell to creat such things. So in addition, I think it would be useful to add a new mode that only executes code on demand.
Nevermind, their is the Breakpoints mode that already do that, Yes.
(however, you must not select such stored template without the Breakpoints mode enable, else your cooked)

[chaley]

Originally Posted by un_pogaz

Else, if retrive the seleted book ids is good, I think it would be much more useful to have a function 'search_books()' that use the search syntax.

ids = search_books('series:"='&field('series')&'"')

rlst = ''
for id in ids:
	with id:
		rlst = rlst & character('newline') & $title
	htiw;
rof;
show_in_dialog(rlst) # show all the books in a series

Use the function book_values(), as in

ids = book_values('id', 'series:"=' & field('series') & '"', ',', '0')

And let's even be bold, adding a 'all_ids_books()', that way, we can run the entire library.

Use this. The null search expression matches everything.

ids = book_values('id', '', ',', '0')

(this two function would also greatly benefit to have a optional argument virtual libray)

The fourth parameter tells book_values() whether or not to use the current VL. If you want to use an arbitrary VL then put it in the search expression. For example, this expression returns the ids of all the books in the 'foo' VL:

ids = book_values(id, 'vl:foo', ',', '0')

But now that I think about it, I don't think the template editor is designed for that kind use, since it executes the code at each edit... it would become a hell to creat such things. So in addition, I think it would be useful to add a new mode that only executes code on demand.
Nevermind, their is the Breakpoints mode that already do that, Yes.
(however, you must not select such stored template without the Breakpoints mode enable, else your cooked)

Yes, editing a template without breakpoints enabled is painful. I always have it on even when I'm not actually using breakpoints.

NB: I'm also adding a way to execute the template tester where it doesn't run the template for all the selected books, ignoring all but the first book. This helps develop and use templates that iterate over selected ids, generating reports.

[mijkala]

I'm using an action chain to clean up my status section for fics and it's been working well, until I noticed one that won't change? Here's the code I have for both

 program:
now = format_date(today(),'iso');
updated = $$#ao3_last_updated;
status = $#status;

	if 
		status=='In-Progress'
		&& days_between(now, updated) ># 183
	then 
		'Dormant'

	elif
		status=='In-Progress'
		&& days_between(now, updated) ># 900
	then 
		'Hiatus'

	else 
		status
	fi

and

program:
now = format_date(today(),'iso');
updated = $$#ao3_last_updated;
status = $#status;

	if
  		status=='Dormant'  
 		&& days_between(now, updated) ># 900  
	then    
		'Hiatus' 

	else 
		status
	fi

They're both set up under 'Status Clean Up' and effecting my custom column #completed_date , but one fic won't change from "In-Progress" after running this chain, even though the last time it was updated was in 2016. Any ideas why this is happening? I had these two combined together, but tried to separate it and see if that was the issue but it didn't fix it

Here's the fic if you need it https://archiveofourown.org/works/82...pters/18891986

[capink]

Originally Posted by mijkala

Would someone be able to help me set this up? I'm new to action chains (I just have the one setup to change status of fic based on time and that was basically copied and pasted - thank you ownedbycats for that). I have no clue how to set this up, but this is exactly what I want to do basically

Since I don't use fanfic, I need to see and example of fanfic identifier and the corresponding url to be able to help. Hopefully other who use this plugin and fanfic can chime in for help as well.

[capink]

Originally Posted by chaley

FWIW: I'm currently implementing these changes to the base formatter.

Thanks chaley, this is going to make it easier and more elegant, and will render some of the functions I've made like selection_count() and from_selection('id') unnecessary, but I will keep them for backward compatibility.

For those interested, once chaley submits his changes, selection_count() can be replaced by a native calibre function as follows:

list_count(selected_books(), ',')

And instead of using from_selection('id') you can use selected_books(). from_selection() can be still used with other columns and can be useful in formulas action.

Originally Posted by un_pogaz

self.chain_selected_ids = self.gui.current_view().get_selected_ids()

This call can be very expensive if the user have a lot of books selected. I already had a similar call in earlier iterations of the plugin (to provide the selection_count as a variable) and had to scrap it because it can slow down chains that does not benefit from it. So this is a non starter for me. I've already tried that path and came out with the conclusion that the call has to be made explicitly by the user only in the chains the might need it, like the solution I linked to in an earlier post. That is also I why I wrote the on-demand selection_count() function instead of providing a chain variable that might not be needed in most of the chains.

As for the rest of the implementation, it can be done in a custom action to avoid cluttering the interface of the plugin with another column. I don't think it provides a better solution to what we already can do in the "Search Using Templates" action, especially with the changes chaley made.

[chaley]

Originally Posted by mijkala

I'm using an action chain to clean up my status section for fics and it's been working well, until I noticed one that won't change? Here's the code I have for both

program:
now = format_date(today(),'iso');
updated = $$#ao3_last_updated;
status = $#status;

	if 
		status=='In-Progress'
		&& days_between(now, updated) ># 183
	then 
		'Dormant'

	elif
		status=='In-Progress'
		&& days_between(now, updated) ># 900
	then 
		'Hiatus'

	else 
		status
	fi

Of note:

• The only check that will succeed is "># 183". Both 500 and 1000 are larger than 183. You should reverse the two checks to look for "># 900" first.
• today() returns an ISO formatted date. You don't need to format it again.

As for why the fic isn't updated, I suspect that #ao3_last_updated doesn't contain what you think or is formatted in some way where the conversion of the string to a date/time is failing. Without seeing the actual metadata in the library for the book we can't say any more. The book itself isn't relevant.