How would you search in TYPO3 Backend?

Discussion Topic

For implementing a continuous user experience throughout the TYPO3 Core, I kindly ask everybody to answer the following three questions.

What would you type into a search field (no matter if within TYPO3 or any other system) in order to retrieve all results including

  1. the words John Doe in this exact phrase?
  2. the words sandstorm and butter in an arbitrary occurrence?
  3. either the word developer and/or integrator?

Impact

We intend to collect responses and will finally use the answers for a voting process, which shall be the guideline on how search syntax should be implemented within the TYPO3 Backend for all search fields.

Organizational

Topic Initiator: Markus Klein
Topic Mentor: Markus Klein

  1. "John Doe"
  2. john +doe or +john +doe
  3. developer integrator

“john doe”
sandstorm+butter
developer integrator

i normally always use lower case

Since you ask, this is what I would prefer (case insensitive):

  1. “john doe” : return only results with exact phrase “John Doe”
  2. +sandstorm +butter (or sandstorm AND butter) : return only results which contain both words
    3 developer integrator : return all results with either developer or integrator, ranking occurences with both higher

Since most people will not be familiar with any search query syntax and the current general trend is simplifying and removing features, I am afraid, you would have to provide best guess results for queries:

  1. john doe : returns all results with john or doe, ranking results with both higher + ranking the phrase higher (or the 2 words in close proximity, see proximity search)
  2. see 1.
  3. see 1.

To be honest:

1: john d
2. sandstorm butter
3. developer integrator

When searching, I mostly don’t care about “and” or “and/or” and I do not complete my search words most of the time, as I expect the search to be fired on-the-fly when typing. As soon as the right result is in the result set, I stop typing. This is true for most of the “everyday searchengines” I use, like Google, PHPStorm or the Ubuntu internal search.

Heck in my OS I usually do not type more then 3 characters to find what I want.

Regarding the integration of a search syntax feature inside of the TYPO3 backend, I would probably only vote for the + sign and the double quotes. The “on-the-fly” searches however are really handy and could be a very good think to have in all search fields in the backend.

  1. “John Doe”
  2. sandstorm butter
  3. developer integrator

I don’t have any sophisticated search fu tbh.

  1. “john doe”
  2. sandstorm butter
  3. developer integrator

“john doe”
sandstorm butter
developer OR integrator

  1. john doe (not before yielding too many results I‘d wrap with double quotes)
  2. sandstorm butter
  3. developer integrator

I rarely miss OR in search syntaxes if it‘s not available. On the other hand, what I find myself doing all the time is panama -canal to find documents with panama that miss canal. Not sure if I‘d need this in TYPO3 BE, though.

knowing some search engines I know some syntax they expect and I would try:

  1. “john doe”
  2. +sandstorm +butter
  3. developer integrator

if i notice these gave the wrong result I might try

  1. sandstorm AND butter
  2. developer OR integrator

but I also would expect that the results of any two word search would give the result in an order::

  1. literal findings (the two words following each other)
  2. texts with both words somewhere
  3. texts with any of the words

the usual problems may occur: is one occurrence in a short text more important than multiple occurrences in a long text. and also the occurrences in different sections/ areas in the text.
And that gives the question: could that break the order and result in a higher ranking of a text with only one finding (but very relevant for that word) above a text with both words finding but low ranking (e,g, only one occurrence in a long text)

  1. “john doe” (of course, with quotes)
  2. sandstorm +butter
  3. developer integrator

I don’t use and/or often though. But the + or - syntax mentioned in this thread might make sense, and I’ve seen other people tweaking their results this way.

1: “John Doe”

2: sandstorm butter
Sandstom AND butter
3: developer OR integrator

There is a difference in how I as a developer will search and how an editor will search. So I will gather responses of some of our clients. And will give the average answers as a separate reaction. For now, my own way of searching:

  1. “John Doe”
  2. sandstorm + butter
  3. developer integrator

I agree that admins/devs/integrators search quite differently than editors, e.g. I especially liked the old “table:uid” search functionality. However instead of thinking about an arbitrary construct, perhabs adhering to a defacto standard might be better including many alternatives, thus I’d suggest the Google syntax (https://moz.com/learn/seo/search-operators) as this is also the way I’d expect it, as an OR search is much more uncommon than an AND search.

So the question “how would I search”:

  1. “John Doe”
  2. sandstorm butter
    3.1) developer OR integrator
    3.2) developer || integrator
    3.3) developer | integrator

(I expect OR, || and | to work, but would use them in this order; if and/or would be operators I won’t expect them to work in lower case, however the equivalents & and | should work in single and double versions).

However I agree to some above that a plain word search should always reveal all results just sorted by relevance, as this is more like the defacto standards of the OS (e.g. Spotlight, Cortana, Android and iOS equivalents).

So the question “how should search be”:

  1. john doe
  2. sandstorm butter
  3. developer integrator
    (either 3 won’t show up results or fallback to OR because there are no results)
  1. “john doe”
  2. sandstorm butter
  3. developer integrator

with a priority order: first the results that contain both words (if any) and then the results with single ones

  1. John Doe
  2. Sandstorm Butter
  3. Two individual queries, one for Developer, one for integrator. If Not allowed: Integrator Developer

Expectation: exact phrase Ranking Higher than just both words. Lowest ranking if only one word is found

Use cases where I need “or” in one query are rare. But I do use exclude Syntax “this -notthat” (without parentheses) sometimes.

  1. "John Doe"
  2. +sandstorm +butter or sandstorm butter (lower priority)
  3. developer integrator

I would prefer sticking to a well know query language like the one introduced by google:

https://developers.google.com/issue-tracker/concepts/search-query-language

There you could also say:

page.title:"John Doe"

This would make it easy to start with search and eliminate the need for an additional interface for advanced users.

  1. John Doe
  2. sandstorm butter
  3. developer integrator

Agreeing with the people above who state that search should be as downright simple as it can be. If this doesn’t cut it, maybe add a few de-facto standard Google-isms like ‘format:’, + and - and quoting for phrase-searches. Anything more advanced should go behind and ‘advanced’ function. Less is more if you ask me.

I see a distinction between ‘search’ and ‘query database’ functionality. For an admin/developer I can imagine a simple box that accepts full-blown SQL SELECT statements or some other powerful algebraic syntax, but that’s not searching… it’s querying and those two should not get mixed.

I just noticed: what about excluding words?
is it a common syntax?
how to search for ‘browser’ but exclude all pages which mention ‘Explorer’?

Thanks everybody for these detailed answers.

Actually we don’t have such sophisticated search mechanisms yet. My intention was more or less to find out whether most of you would quote an exact phrase or not.

Keep in mind that we do not talk about a specific search field here, but about all search fields in the Core like Live-Search, Filelist, Suggest-Wizard.

I will set up a voting soon.