There are some Format* functions in timetrace.go:

Those functions should be part of the Formatter type in formatter.go. Once those functions are moved, they can then be called with t.Formatter().FormatXxx from the cli package.

Hey,

I am currently working on multiple subprojects and want to use Timetrace to track my work time on these.

My initial idea was to do something like this : parentProject/SubProject, but it confuses Timetrace 😄

Cheers

Similar to issue #81 , the timetrace edit record <key> command does not rename the file. This shouldn't be an issue when we pass either the --plus or --minus flags, since these edit the end times, but without those flags, the user can edit the json file and potentially change the start time, which is the key for a record. The documentation recommends using timetrace list records in conjunction with the timetrace edit record command, but this does not work as intended if a flag is not used.

The code expects the key to correspond to the filename, but since it is not changed, the file is not found.

In the image, I demonstrate the above with record 1. Note that this also occurs when use12Hours is false.

I'd imagine the solution would be similar to those mentioned in issue #81 .

Instead of writing timetrace list records 2021-05-17 every time, I'd like to be able to just write timetrace list records today.

This means there should be aliases called today and yesterday that translate those arguments to the respective dates.

Another thing that just came to my mind regarding #72 and #53 is that delete project should probably also delete all submodules of the project, right? Otherwise we could again run into problems with "parentless"-modules.

Support for Clockify would be great. There exists an API for it.

Hi,

Nice tool! Maybe I didn't RTFM correctly but I can't find a way to display the total time spent on a single project ?

Is this implemented or will it be ?

Thanks,

The timetrace list records command should have a --project <KEY> filter flag to only display records for the given project - similar to --billable.

At the moment, running timetrace status when there's no active time tracking will return an error:

timetrace status
❗ Failed to obtain status: start tracking first

This shouldn't be an error. Maybe out.Info would be a better fit here.

Currently, creating project modules like grind-beans@make-coffee is possible even if the parent project, make-coffee, doesn't exist. This shouldn't be allowed and throw an error instead.

More information about project modules and their design can be found in proposal #53.

Just like a record can be displayed using timetrace get record <YYYY-MM-DD-HH-MM>, a record should be deletable using timetrace delete record <YYYY-MM-DD-HH-MM>.

This means that the delete command should get a new sub-command called record, which takes the record date as argument - and there needs to be a DeleteRecord method of core.Timetrace.

The documentation needs to be updated as well.

Running timetrace edit record latest when a user has not logged any time today causes the program to panic, image attached. Note, I did have valid records from the day prior, so I imagine it should have grabbed the record from yesterday, but it looks like the code assumes the newest directory in ~/.timetrace/records will contain time logs.

I think the output of timetrace list records YYYY-MM-DD would be more useful if the project name was included in the output table

Current output of timetrace list records YYYY-MM-DD:

Desired output:

At the moment, the output of timetrace status looks as follows:

+-----------------+--------------------+-----------------+
| CURRENT PROJECT | WORKED SINCE START |  WORKED TODAY   |
+-----------------+--------------------+-----------------+
| make-coffee     | 8h25m26.147239s    | 8h25m26.147239s |
+-----------------+--------------------+-----------------+

Displaying durations like 8h25m26.147239s is precise, but unnecessary und unreadable. It should rather look something like 8h 25min.

To keep dependencies up-to-date, Dependabot should be configured.

I'd suggest a configuration file similar to the one of verless.

Currently, running go run . edit record will panic in case the latest record has not been stopped yet - because in that case, the record's End field is nil. Editing a record using the --plus and --minus options should not be allowed if the latest record has no end yet.

There should be a new --revert flag for timetrace edit project, timetrace edit record (once it is implemented through #19), timetrace delete project and timetrace delete record.

If the --revert flag is specified, the edited/deleted project/record should be restored. This means that the edit and delete commands also need to create a backup file before modifications are performed. There always should be one backup file per project/record.

Those backup files must not be taken into account by functions like ListRecords.

If the --revert flag is specified, nothing except the rollback should happen.

Affected commands

¹ This command doesn't exist yet.

I'm planning for a timetrace report command which should generate a report of the tracked time. This issue is a small idea collection for the new command.

@FelixTheodor, @rocar, @rknizzle, @aaronsheah, @joshuaherrera, @Gealber, @KonstantinGasser, @adzo261, @obnoxiousnerd, @aligator and @nrobinson2000 should feel free to add their suggestions here!

Since the editing of records will probably often used for the last stopped record, I would propose to introduce an alias in a similar way as #30 for edit record .

If you like the idea @dominikbraun , you can assign me this issue :)

At some places in the code base, there is the hard-coded time string "15:04". This should be replaced with a defaultTime constant which should be used instead. Plus, if Timetrace.Config().Use12Hours is true, that default constant should be replaced with 03:04PM.

Hi,
After creating a project as per README usage instructions, whenever I try to start tracking I get this error on Mac:
$ timetrace start coding
❗ Failed to start tracking: start tracking first

Please advise. Thank you!!

It would be nice to filter the listed records so that only billable records will be displayed:

timetrace list records --billable
# Only lists billable records

I just stumbled across this project: https://github.com/gen2brain/beeep
It makes sending notifications to the user very easy. Then I had an idea - my main concern with time-tracking software is that I always forget the tracking when I'm in the middle of something.
Therefore, it could be a useful feature for timetrace if the user can set a timer to get a notification when the current task he is tracking is running for x minutes.

I already implemented a first version of this and opened a WIP-Pull-Request for it.

You can start a record and then run timetrace notify 5m & (the &-symbol makes the process run in background). As soon as the task is running for 5 minutes, a notification should pop up and the notify process finishes. If the tracking is stopped before 5 minutes are over, no notification should be send. On my linux system, this already works fine.

What do you think of this idea and the first implementation, @dominikbraun ?

Project modules, also known as sub-projects, are a frequently requested feature. This is my proposal on how to implement them.

Proposal

The idea

Project modules are the exact same thing as projects, expect that they can be associated with a parent project. This association is marked by an @. For example, grind-beans@make-coffee is a module of the project make-coffee.

Creating a project module

The syntax is the same as for creating a "normal" project. The logic for creating the project remains unchanged - it will simply create a file called [email protected] internally.

timetrace create project grind-beans@make-coffee

Tracking time for a project module

The start logic remains unchanged as well.

timetrace start grind-beans@make-coffee

Filtering records for a module

Once #52 is implemented, the list records command has a --project flag for filtering projects. With project modules, the respective logic would have to take an @ into account and additionally filter by module.

timetrace list records --project grind-beans@make-coffee today

Implementation

As mentioned above, project modules are stored in the exact same way as projects. Our example module will be stored as [email protected] inside the projects directory.

To list all available modules of the project make-coffee, all "projects" - e.g. modules - suffixed with @make-coffee can be displayed.

The goal of #84 was to find a way to auto-generate documentation out of the Cobra commands. @obnoxiousnerd found a simple way to do this natively with Cobra. Thanks!

As a next step, the command references from README.md should me copied into the Cobra commands. Each Cobra command has a Long field for long help texts, and the existing documentation should be moved into that field.

Ideally, as a result, generating the documentation from the commands should produce the same output as the current Markdown documentation in README.md.

timetrace list records lists all records of a given date. I think it would make sense to summarize the total tracked time at the bottom of the table.

The table should look like this: https://github.com/olekukonko/tablewriter#example-2---without-border--footer--bulk-append

Regarding Proposal #53:

The timetrace list projects command shall not only list projects and project modules row-by-row, e.g.:

Instead, modules should be listed as part of their respective project:

This should be relatively easy to implement - just create a function that loads all modules of a project and display them with the project.

Running timetrace status leads to an error if time hasn't been tracked on that day yet:

timetrace status
❗ Failed to obtain status: open /Users/dominik/.timetrace/records/2021-05-15: no such file or directory

This is because the record directory of that day, in this case 2021-05-15, hasn't been created yet. It will be created when running timetrace start.

To fix this, Fs.EnsureDirectories should also create a record directory for the current day:

EnsureDirectories is executed in advance on each timetrace command.

Description:

If there is no active time tracking, timetrace stop still prints a success message that tracking has been stopped.

Steps to reproduce:

1. Start tracking time.
2. Stop tracking time.
3. Stop tracking time again: ✔️ Stopped tracking time

Expected behavior:

If there is no active time tracking, stop should do nothing. There should be an info (probably not an error?) message indicating that time isn't being tracked at the moment:

There is no active time tracking. Start tracking using `timetrace start` first.

If you follow the instructions:

sudo snap install timetrace --edge

error: The publisher of snap "timetrace" has indicated that they do not consider this revision to
       be of production quality and that it is only meant for development or testing at this point.
       As a consequence this snap will not refresh automatically and may perform arbitrary system
       changes outside of the security sandbox snaps are generally confined to, which may put your
       system at risk.

       If you understand and want to proceed repeat the command including --devmode; if instead you
       want to install the snap forcing it into strict confinement repeat the command including
       --jailmode.

Info:

• snap 2.50
• snapd 2.50
• series 16
• ubuntu 18.04
• kernel 4.15.0-143-generic

The timetrace status command prints how long I've been working today. It would be nice to see how long I've been taking breaks since the beginning of my day. For example:

timetrace status
+-------------------+----------------------+----------------+--------+
|  CURRENT PROJECT  |  WORKED SINCE START  |  WORKED TODAY  | BREAKS |
+-------------------+----------------------+----------------+--------+
| web-store         | 0h 3min              | 4h 15min       | 30min  |
+-------------------+----------------------+----------------+--------+

Hi,
In working on issue #17 in version 0.4.0, I noticed that, when 12 hour time format is configured with config.yaml, one must always pass in the PM portion of the flag, even if the time logged was in the AM, as shown.

It seems to me the command should allow for the user to input AM or PM, or perhaps allow the user to omit the PM entirely, with the assumption that any time entered without PM is assumed to have been logged in the AM. Thoughts?

For example, the get record command has the following syntax:

Actually, YYYY-MM-DD-HH-MM is only the default syntax. If the user has set use12hours: true in their config, the expected format is YYYY-MM-DD-HH-MMPM instead, as you can see here:

The Use field of the command should be changed accordingly, being sensitive to the actual configuration.

Hi there,

thanks for the project. It looks nice.
However I encountered a bug trying to stop tracking:

timetrace create project test
✔️ Created project test
timetrace start test
✔️ Started tracking time
timetrace stop
❗ Failed to stop tracking: record already exists

timetrace status works fine.

The json looks like this:

{
	"start": "2021-05-26T10:23:37.5685406+02:00",
	"end": null,
	"project": {
		"key": "test"
	},
	"is_billable": false
}

I use timetrace version 0.6.0 on Windows 10.

Just as timetrace list projects displays a list of projects, timetrace list records should list records. It probably would make sense to limit those records to a specific date, so that the command syntax looks as follows:

timetrace list records <YYYY-MM-DD>

The output should be a table analogous to list projects.

Hey, thanks for this awesome tool! Makes the time tracking part of my life easier 👍

Right now I have the current task that I am working on in my status bar.

timetrace status | awk -F'|' 'NR == 4 {gsub(/ /, "", $0); print $2 " " $3}'

And this works, but it would be nice to add a format parameter to timetrace status to fx format the output as JSON, which is easier to work with in scripts :-)

The table displayed by timetrace list records should contain a new column "key". This should be the "record key" in the form YYYY-MM-DD-HH-MM - i.e. the exact value that you would pass to timetrace get record to display that record.

The bug

timetrace status uses <current time> - <start time of the last record> to determine the worked time. This is correct if the time is being tracked, but incorrect when tracking has been stopped. In that case, <end time of the record> - <start time of the record> would be correct.

As a result, if you run timetrace status multiple times after stopping time tracking, it looks like time is still being tracked.

Steps to reproduce

1. Run timetrace start ....
2. Wait.
3. Run timetrace stop.
4. Run timetrace status.
5. Wait.
6. Run timetrace status again.
7. It looks like the time is still being tracked, because the tracked time between the first and the second status command increased - even though tracking has been stopped before.

Expected behavior

After running timetrace stop, the worked time for the last record displayed with status shouldn't change anymore.

Thanks to tablewriter, tables already look pretty good! However, the out.Table function could be improved: tablewriter allows to stylize and colorize tables. I think it would be nice to have at least bold table headings.

To do this, table.SetHeaderColor needs to be called here:

table.SetHeaderColor expects one function argument per table column. This has to be done automatically, depending of the length of header.

Just as projects are editable via timetrace edit project, records should be editable using timetrace edit record. The command syntax should look just like the syntax of timetrace get record:

timetrace edit record [options] <YYYY-MM-DD-HH-MM>

• Running edit without any options should open the default editor to directly edit the record file (see edit project).
• Running edit with options should load the record internally, apply the option values and save the record. The supported options are explained below.

Command options

Those options probably aren't too hard to implement if the durations can be parsed as time.Duration. In that case, calling something like record.End.Add(plus) should be sufficient.

At the moment, to get and set configuration values, users have to create a config.yaml file inside their .timetrace directory and set the values manually.

It would be nice to have a timetrace config command with two sub-commands to do this:

• timetrace config set <KEY> <VALUE>
• timetrace config get <KEY>

This way, users wouldn't have to touch their configuration file anymore.

Just a suggestion, we could move to use a more interactive cli library. This might be useful for example where you can run timetrace start without arguments and you can use arrow keys to select the project you would like to start.

What do you think? I can come up with a proof of concept first before we decide.

Running the create project <key> is not checking for existing projects correctly, even though there is code to handle this case.
I'm running Manjaro Linux, in case this is an issue with how the os package interacts with linux, and I tested on timetrace 0.5.0.

Per the code, this would also effect the record.SaveRecord function.

At the moment, the output of timetrace version looks as follows:

$ timetrace version
❗ timetrace version v0.1.0

This error-styled output should be replaced with an info-styled output, which can be accomplished by replacing out.Err with out.Info:

When running the timetrace delete record, the record to delete should be displayed (just as get record does), and the user should be asked for confirmation with a yes/no question.

The confirmation should be skippable by providing a --yes flag.

Sometimes, it would make sense to directly "insert" a new record with a given time span instead of fiddling around with timetrace edit record on existing records.

To do so, a timetrace create record command is needed. I have the following syntax in mind:

timetrace create record <PROJECT KEY> {<YYYY-MM-DD>|today|yesterday} <HH-MM> <HH-MM>

Important: This command should check for collisions with other records, so probably all records from that date have to be loaded and checked internally.

Currently, when a user edits a project using timetrace edit project <key> the text editor is opened so the user can change the key.
It seems to me that any subsequent calls to this command should reference the new key, but trying to run this command this way causes the file to not be found.

Since some functions like project.LoadProject assume that the key will be part of the path, it seems like this command should also rename the json file with the new key.

I'd love to auto-generate the documentation for commands from the cobra.Command.Long help text and the command's options. Maybe there's some cool way to do this?

Currently, tables look as follows:

Those headers are nice, but there should be a padding so that they look like so:

This can be done by prepending and appending a whitespace before and each header name: "Project" should become " Project ". This modification should be performed by out.Table:

Just create a function headersWithPadding(headers []string) []string and use it in out.Table.