Restrict directory names

[thorehusfeldt]

In 2023-07-draft, we have

All file names for files included in the package must match the regexp
^[a-zA-Z0-9][a-zA-Z0-9_.-]{0,253}[a-zA-Z0-9]$
i.e., they must be of length at least 2, at most 255, consist solely of lower- or uppercase letters a–z, A–Z, digits 0–9, period, dash, or underscore, but must not begin or end with a period, dash, or underscore.

This allows 1.in, aa and foo.bar.baz_baz-baz but disallows 1_. Fine so far.

For directories, there is no restriction, as far as I can tell. Thus the following are all valid:

• the testgroup secret/1/,
• the testcase input file secret/1/1.in
• the input validator input_validators/v (containing input_validators/v/validate.cpp and input_validators/v/validate.

and so on. I think all of these are fine.

What is also currently allowed, as far as I can tell are

• the testgroup secret/_/,
• the testgroup secret/huge.connected,
• the testcase input file secret/xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx/1.in,
• the input validator input_validators/[﷽,
• the submission directory submissions/😊/th.py,

as well as several other references to validators, testgroups, testcases, visualisers or submissions that use “whatever the operating system supports”.

I propose to tighten the specification as follows:

All names for directories included in the package must match the regexp
^([a-zA-Z0-9][a-zA-Z0-9_-]{0,253}[a-zA-Z0-9]|[a-zA-Z0-9])$
i.e., they must be of length at least 1, at most 255, consist solely of lower- or uppercase letters a–z, A–Z, digits 0–9, dash, or underscore (no period), but must not begin or end with a period, dash, or underscore.

Moreover, I would like to clarify that

Paths that appear in yaml files (including the names of test cases, test groups, or submissions) use the forward slash / to separate names.

[RagnarGrootKoerkamp]

sgtm. Regex would be a bit shorter as (spaces for clarity)

^[a-zA-Z0-9] ( [a-zA-Z0-9_-]{0,253} [a-zA-Z0-9] )? $

[eldering]

Sounds good to me!

[niemela]

Overall, sounds very good. Some minor comments below.

In 2024-08, we have

What are you referring to? We don't have a 2024-08 specified anywhere?

I propose to tighten the specification as follows:

All names for directories included in the package must match the regexp
^([a-zA-Z0-9][a-zA-Z0-9_-]{0,253}[a-zA-Z0-9]|[a-zA-Z0-9])$
i.e., they must be of length at least 1, at most 255, consist solely of lower- or uppercase letters a–z, A–Z, digits 0–9, dash, or underscore (no period), but must not begin or end with a period, dash, or underscore.

Why not use the same for files and directories?

I wonder why we felt that filenames must be of length at least 2?

Moreover, I would like to clarify that

Paths that appear in yaml files (including the names of test cases, test groups, or submissions) use the forward slash / to separate names.

👍

[thorehusfeldt]

What are you referring to? We don't have a 2024-08 specified anywhere?

My mistake. Fixed.

Why not use the same for files and directories?

One reason is .. For files, we do need to be able to express 1.in. On the other hand, for directories, the full stop leads to problems, for instance there are traditions such as YAML where dot has special meaning. In particular, secret.foo.bar is the same as

secret:
   foo:
      bar:

So if we allow a directory named data/secret/foo.bar/ but also data/secret/foo/bar then I can see all kinds of weird problems occurring.

I wonder why we felt that filenames must be of length at least 2?

No idea, but I like it.

[thorehusfeldt]

As an extra gift, somewhat related, I would really-really like to forbid:

data/secret/huge/1.in
data/secret/huge.in

i.e., forbid having a test group and a test case in the same directory (in this case data/secret) and the same name (in this case, secret/huge).

[RagnarGrootKoerkamp]

As an extra gift, somewhat related, I would really-really like to forbid ...

Agreed; let's not allow that.

secret.foo.bar is the same as ...

That's not actually the case in YAML itself, but there are some frameworks that automatically translate this into secret: {foo: {bar: ...}}.
Either way, not having periods in directory names sgtm.

[niemela]

@thorehusfeldt it seems we all agree on this. Do you feel up to writing a PR for this?

[niemela]

@thorehusfeldt it seems we all agree on this. Do you feel up to writing a PR for this?

N.v.m., I created a PR.

[Tagl]

This is done, right?

[eldering]

This is done, right?

Yes, in #357