Postulo/postulo

Fork 0

Company sectors are chosen from a real vocabulary, as labels #141

New issue

Closed

opened 2026-09-09 10:41:57 +00:00 by tiagoagueda · 1 comment

tiagoagueda commented

2026-09-09 10:41:57 +00:00

Owner

Observation

company sectores seams incomplete, you shoud looke for a more extenside and general area of
activity list, and the selection should be in the form of labels, not tick boxes

What exists

Two halves, and the complaint lands on both.

The vocabulary. jobs/industries.py holds 32 hand-written names offered as suggestions;
the real list is Industry, one vocabulary per person, created on the fly by slug. The model
docstring says why it is per person: "two people on one instance may describe the same
employer differently, and neither should see the other's words." A company may belong to
several — "a bank that is also an insurer and a software house is all three, and counting it
under one would be a lie the figures repeat."

The control. CompanyForm offers those industries as a CheckboxSelectMultiple, with a
free-text Other industries box beside it and a <datalist> of suggestions behind that.
Two controls for one idea: tick the ones you have, type the ones you do not.

What this asks for

A vocabulary worth picking from, and a picker made of labels.

Worth being careful about

The two halves are separable and should be done in either order, which is why they are
filed apart. The label control is worth building whatever list is chosen, and it is wanted on
application tags too. The vocabulary decision is worth making whatever the control looks
like.

Two controls become one, and that is most of the interface work. With labels there is no
reason for Other industries to be a separate box: typing a name that does not exist is the
same gesture as choosing one that does. Removing the second field also removes the moment
where somebody ticks three boxes, types a fourth name, and cannot see the four together.

The count on a company is not small. A bank that is also an insurer and a software house
is three labels; a conglomerate is more. The control has to look right at one, at three and
at twelve, and the companies table shows industries in a cell that is already narrow.

Filtering by industry is where a longer list is felt. The companies table filters by
industry as a choice column; 88 divisions in a select is a different control from 32, and
the Industries page — which lists them, renames them, merges them — becomes a page about a
much longer list.

Merging exists and gets more use, not less. Every company in this industry moves to the
selected one, and this one disappears is already there, and it is exactly the tool somebody
needs after seeding a standard list beside their own words. It should be easy to reach from
wherever the seeding happens.

Nothing should rename what a person already wrote. Their industries are their words; a
new vocabulary is offered beside them, never instead of them, and never silently.

Every widget that counts by industry inherits the change — the industries dashboard
widget, the insights breakdown, the same figures by field on a company. A longer list makes
those sparser, and a company in three of them still counts in three.

## Observation > company sectores seams incomplete, you shoud looke for a more extenside and general area of > activity list, and the selection should be in the form of labels, not tick boxes ## What exists Two halves, and the complaint lands on both. **The vocabulary.** `jobs/industries.py` holds 32 hand-written names offered as suggestions; the real list is `Industry`, one vocabulary per person, created on the fly by slug. The model docstring says why it is per person: *"two people on one instance may describe the same employer differently, and neither should see the other's words."* A company may belong to several — *"a bank that is also an insurer and a software house is all three, and counting it under one would be a lie the figures repeat."* **The control.** `CompanyForm` offers those industries as a `CheckboxSelectMultiple`, with a free-text *Other industries* box beside it and a `<datalist>` of suggestions behind that. Two controls for one idea: tick the ones you have, type the ones you do not. ## What this asks for A vocabulary worth picking from, and a picker made of labels. ## Worth being careful about **The two halves are separable and should be done in either order**, which is why they are filed apart. The label control is worth building whatever list is chosen, and it is wanted on application tags too. The vocabulary decision is worth making whatever the control looks like. **Two controls become one, and that is most of the interface work.** With labels there is no reason for *Other industries* to be a separate box: typing a name that does not exist is the same gesture as choosing one that does. Removing the second field also removes the moment where somebody ticks three boxes, types a fourth name, and cannot see the four together. **The count on a company is not small.** A bank that is also an insurer and a software house is three labels; a conglomerate is more. The control has to look right at one, at three and at twelve, and the companies table shows industries in a cell that is already narrow. **Filtering by industry is where a longer list is felt.** The companies table filters by industry as a `choice` column; 88 divisions in a select is a different control from 32, and the *Industries* page — which lists them, renames them, merges them — becomes a page about a much longer list. **Merging exists and gets more use, not less.** *Every company in this industry moves to the selected one, and this one disappears* is already there, and it is exactly the tool somebody needs after seeding a standard list beside their own words. It should be easy to reach from wherever the seeding happens. **Nothing should rename what a person already wrote.** Their industries are their words; a new vocabulary is offered beside them, never instead of them, and never silently. **Every widget that counts by industry inherits the change** — the industries dashboard widget, the insights breakdown, *the same figures by field* on a company. A longer list makes those sparser, and a company in three of them still counts in three.

tiagoagueda added this to the 0.3.0 milestone

2026-09-09 10:41:57 +00:00

tiagoagueda added the

tier

enhancement

interface

labels

2026-09-09 10:41:57 +00:00

tiagoagueda added a new dependency

2026-09-09 10:41:58 +00:00

#139 A control for choosing several labels, without a mouse and without a script

tiagoagueda added a new dependency

2026-09-09 10:41:58 +00:00

#140 Where the list of areas of activity comes from

tiagoagueda added

tier/4

and removed

tier

labels

2026-09-09 11:49:18 +00:00

tiagoagueda referenced this issue from a commit

2026-09-10 03:04:49 +00:00

Make a long vocabulary of sectors usable where it is met

tiagoagueda commented

2026-09-10 03:05:16 +00:00

Author

Owner

Done in e3a4d9aa, and this is where the two halves meet.

The vocabulary is NACE Rev. 2.1 at division level, 87 names in 24 official languages, CC BY 4.0 (#140).
The picker is chips, layered over the controls that already worked, with a name that does not exist yet visible as new before it is saved (#139).

You filed them apart and said they could go in either order; they did, and what was left for this issue is the part that only shows up once both are true.

Two controls became one

With labels there is no reason for Other industries to be a separate box, and it is not one any more: typing a name that does not exist is the same gesture as choosing one that does, and all of them are visible together as chips. The moment you named — ticks three boxes, types a fourth name, cannot see the four together — is gone.

With scripts off the two native controls are still there and still work, which is the honest fallback rather than a lesser one.

The count on a company is not small

The table draws industries as labels now: four shown, the rest counted. A bank that is also an insurer and a software house is three; a conglomerate is more; and a cell that grows without limit pushes every other column off the screen. Tested at one, at three, at twelve, and at none.

Filtering, which is where a longer list is felt

You wrote that 88 divisions in a select is a different control from 32 — and it is not a select. The industry column filters as text against the names, which is the same control at thirty-two names and at three hundred and the only one that stays usable at both. That is a decision confirmed and now held by a test, rather than one changed.

The Industries page, at length

a search that matches the name or the NACE code, so somebody handed a code by an employment office can type the code;
the code shown beside the person's own word for it, which is what the standard bought beyond coverage;
Edit or merge rather than Edit, because merging is exactly the tool somebody needs after picking from a standard list beside their own words, and it should not be hidden behind a word that does not mention it;
the box appears only past a handful of rows — a control above five rows is a control asking to be ignored.

And the page does not become long by itself. A row exists only once a company has been given that industry: the classification is a list of suggestions, not a list of rows. That is the property which made it safe to make the suggestions long, and it is a test now rather than an assumption.

The promises kept

Nothing renames what somebody already wrote. Fintech stays Fintech and carries no code, and there is a test saying no classification is asserted over it.
A company in three fields still counts in three — the widget, the insights breakdown and the same figures by field all inherit the longer list without inheriting a different rule.

Also

tests/test_sectors.py, 16 tests. Suite 4441 passed, 29 skipped; browser suite 72 passed.
Five new strings, filled in all 39 European catalogues.

Done in `e3a4d9aa`, and this is where the two halves meet. - **The vocabulary** is NACE Rev. 2.1 at division level, 87 names in 24 official languages, CC BY 4.0 (#140). - **The picker** is chips, layered over the controls that already worked, with a name that does not exist yet visible as new before it is saved (#139). You filed them apart and said they could go in either order; they did, and what was left for this issue is the part that only shows up once both are true. ## Two controls became one With labels there is no reason for *Other industries* to be a separate box, and it is not one any more: typing a name that does not exist is the same gesture as choosing one that does, and all of them are visible together as chips. The moment you named — *ticks three boxes, types a fourth name, cannot see the four together* — is gone. With scripts off the two native controls are still there and still work, which is the honest fallback rather than a lesser one. ## The count on a company is not small The table draws industries as labels now: **four shown, the rest counted**. A bank that is also an insurer and a software house is three; a conglomerate is more; and a cell that grows without limit pushes every other column off the screen. Tested at one, at three, at twelve, and at none. ## Filtering, which is where a longer list is felt You wrote that *88 divisions in a select is a different control from 32* — and it is not a select. The industry column filters as **text** against the names, which is the same control at thirty-two names and at three hundred and the only one that stays usable at both. That is a decision confirmed and now held by a test, rather than one changed. ## The Industries page, at length - a **search** that matches the name *or* the NACE code, so somebody handed a code by an employment office can type the code; - the **code shown beside the person's own word for it**, which is what the standard bought beyond coverage; - ***Edit or merge*** rather than *Edit*, because merging is exactly the tool somebody needs after picking from a standard list beside their own words, and it should not be hidden behind a word that does not mention it; - the box appears only past a handful of rows — a control above five rows is a control asking to be ignored. **And the page does not become long by itself.** A row exists only once a company has been given that industry: the classification is a list of *suggestions*, not a list of rows. That is the property which made it safe to make the suggestions long, and it is a test now rather than an assumption. ## The promises kept - **Nothing renames what somebody already wrote.** *Fintech* stays *Fintech* and carries no code, and there is a test saying no classification is asserted over it. - **A company in three fields still counts in three** — the widget, the insights breakdown and *the same figures by field* all inherit the longer list without inheriting a different rule. ## Also - `tests/test_sectors.py`, 16 tests. Suite 4441 passed, 29 skipped; browser suite 72 passed. - Five new strings, filled in all 39 European catalogues.

tiagoagueda closed this issue

2026-09-10 03:05:17 +00:00

tiagoagueda referenced this issue

2026-09-18 10:12:46 +00:00

Five list and grid pages still sit in the 1280-pixel column: the report, tags, industries, an application's documents and the import mapping #273