Maintenance refactor - 2nd pass by AESternberg · Pull Request #72 · jdeath/CheckRoyalCaribbeanPrice

AESternberg · 2026-05-29T03:56:26Z

Please note the Pending Questions section at the bottom. These were things that I came across but didn't want to change without your input or decision.

Description

This PR introduces an architectural modernization of the cruise tracker script focused on DRY (Don't Repeat Yourself) principles, rigid type-safety boundaries, clean multi-path request unification, and enhanced runtime resilience against unstable remote API responses.

Key Enhancements & Structural Themes

DRY Code Consolidation (_execute_api_request): Centralized duplicate network communication layers, header assembly mechanics (Access-Token, vds-id, AppKey), and validation routines into a single request engine. This eliminates redundant boilerplate code across downstream querying components.
Unified Multi-Path Execution (Path A & Path B Alignment): Standardized how automated reservations are processed by leveraging _build_checkout_url to synthesize clean link strings. Funneling this output back into parse_provided_URL unifies both Path A and Path B requests into an identical, predictable stream before hitting the downstream pricing modules.
Strict Domain Type Hardening: Implemented comprehensive type annotations across all primary routing hooks and utility managers. Generic tracking dictionaries have been systematically refactored to consume explicit structural interfaces and models (AccountInfo, CruiseURLParams, DiscountProfile, WatchListItem).
Defensive Boundary Execution (get_room_price_via_API): Enclosed deep multi-layered JSON payload traversals (such as Next.js RSC string processing and response.json().get("rooms") nodes) inside specialized try/except scopes. This safeguards the master script loop from unexpected root-level dictionary or list structure shifts during server-side modifications or outages.
Robust Input Normalization (get_final_payment_date): Standardized dynamic deadline offsets (75/90/120 day rules) by implementing a sanitization interface capable of translating raw text representations, arbitrary dash/slash timestamps, and standard datetime objects interchangeably.
Technical Debt & Scope Leak Remediations:
-- Fixed a casing typo (self.all_Included = False) inside apply_overrides that accidentally mutated an unmapped property instead of updating the actual dataclass tracker flag.
-- Cleaned up redundant buffering infrastructure (self._buffer) from the PrintRedirector log class.
-- Standardized local variables, dictionary keys, and function configurations across the entire script to meet uniform PEP 8 snake_case patterns.

Pending Questions (Technical Debt)

I noticed two sets of ANSI color codes between this script and Browse*.py. For better maintainability, I recommend picking one for commonality.
I noticed a number of the API servers are hard-coded to use "royal" even though a cruise may be with Celebrity (for example, aws-prd.api.rccl.com/en/royal/web/commerce-api/catalog in get_new_beverage_price. Is that required, or should they go to the correct cruise line URL?
Speaking of get_new_beverage_price, it seems to me that this function is used for all add-ons, not just beverages. If so, I recommend renaming the function to reflect that, as it's better for understanding and maintenance
Is it possible to have a Club Royale ID without having a C&A number? If not, I propose moving that check under the C&A check (again, for streamlining and maintainability). Similarly for Blue Chip and Captain's Club
I want to understand your design: is it the intent that all guests would have the same stateroom subtype and category code in calculate_passenger_metrics? I think so, but wanted to confirm
Finally, I noticed all the passenger discount qualifiers were hard-coded to 'n' when translating to the various r0* fields in check_if_room_is_available? Is that by design, or should those values use what's read in from the config YAML?

…beanPrice

jdeath · 2026-05-31T19:41:52Z

Still have at least a couple bugs:

XXXX Brilliance of the Seas BALCONY CB (Residency Discount) Not For Sale
Sent ntfy notification to 'https://ntfy.sh'.
        Available Rooms (non-discounted price) for 2 Adult and 0 Child on This Sailing Are:
ERROR: NameError: name 'newKey' is not defined

Fixed that, I immediately found another:

yalcaribbean\CheckRoyalCaribbeanPriceAES.py", line 1672, in get_orders
    reservations=getattr(watchItem, 'reservations', [])
                         ^^^^^^^^^
NameError: name 'watchItem' is not defined

I do not even have watch items. I can't imagine why that would not have been found in testing.

First, if booked cruise, should not get a notification if no longer on sale. I only notify if a watch list cruise is no longer available.

For your other questions:
I noticed two sets of ANSI color codes between this script and Browse*.py. For better maintainability, I recommend picking one for commonality.
Use the Check script ones. though I like the blue in the browse*

I noticed a number of the API servers are hard-coded to use "royal" even though a cruise may be with Celebrity (for example, aws-prd.api.rccl.com/en/royal/web/commerce-api/catalog in get_new_beverage_price. Is that required, or should they go to the correct cruise line URL?

I do not know. I do not have celbrite cruises, so do not know the URL. I bet they both resolve to same place, so no need to alter.

Speaking of get_new_beverage_price, it seems to me that this function is used for all add-ons, not just beverages. If so, I recommend renaming the function to reflect that, as it's better for understanding and maintenance

That is fine. it was only beverage when I first wrote it. it does alot more now

Is it possible to have a Club Royale ID without having a C&A number? If not, I propose moving that check under the C&A check (again, for streamlining and maintainability). Similarly for Blue Chip and Captain's Club

I do not think so. I think you need a C&A before you have a casino. same with celeb

I want to understand your design: is it the intent that all guests would have the same stateroom subtype and category code in calculate_passenger_metrics? I think so, but wanted to confirm
** I think so**, if they are in the same room.

Finally, I noticed all the passenger discount qualifiers were hard-coded to 'n' when translating to the various r0* fields in check_if_room_is_available? Is that by design, or should those values use what's read in from the config YAML?
** They should default to no, unless overridden by the account or the config.yaml . Config.yaml should override what is in the account**

AESternberg · 2026-05-31T20:57:02Z

I have a few watch items and they were coming up OK; it's likely that not having them was another execution path I hadn't taken. I'm not surprised: what you found are some camelCase to snake_case conversions that were unfortunately missed (I corrected them manually when refactoring, so it's hard to catch them all, especially sincethe server keys are in camelCase and have to be kept like that). They're being flagged even without watch items because of how the Python interpreter parses the script (which I don't fully understand myself).

Thanks for answering the questions; I'll make those updates and submit an update.

AESternberg · 2026-05-31T21:34:37Z

I'm still making changes based on your answers to my questions (thank you again!), but still had questions on the colors.

As background, I was looking to be consistent in the codes:

\033[<code>m means "Bright text, default background, normal weight", where code is 91-94
\033[1;<code>;40m means "Bold standard text, black background", where code is 31-34.

The currently used definitions are (you see that RED follows a different standard):
RED = '\033[1;31;40m'
GREEN = '\033[1;32m'
YELLOW = '\033[33m'
BLUE = '\033[94m' (from Browse*.py)

Here's how Google describes the differences:
Code Breakdown

\033[9m: High-intensity (bright) color text with default background.
\033[1;3;40m: Bold standard color text with a black background.

Detailed Parameter Comparison
ANSI escape sequences use numbers separated by semicolons to apply styles. Here is what each specific number inside your codes triggers:

Component	\033[9m	\033[1;3;40m
1	Not included (Normal weight)	Bold or increased intensity
Foreground Color	9 (Bright Color)	3 (Standard Color)
Background Color	Not included (Default terminal background)	40 (Black background)

Key Practical Differences

Visual Appearance: \033[9m uses the terminal's native "bright " color slot. \033[1;3;40m uses the "standard" color slot but renders it bold, while forcing the background behind that specific text to turn pitch black.
Terminal Backgrounds: If a user has a transparent or white terminal theme, \033[9m will blend in cleanly. \033[1;3;40m will draw an abrupt black box behind the text.
Compatibility: 9 belongs to the AIXterm high-intensity extension. It is widely supported by modern terminals but skipped by very old legacy systems. 1, 3, and 40 belong to the strict baseline ANSI standard and work everywhere.

So, which do you prefer, [9m style, [1;3;40m style, or mixed as it currently is?

jdeath · 2026-05-31T21:41:50Z

I am fine with mix, not so big on conforming to standards. Better yet, define a config.yaml OPTIONAL override for the colors.

I run in home assistant and the terminal colors took a lot of trial and error to get them to display correctly on the default white background in home assistant logs.

AESternberg · 2026-05-31T21:47:24Z

How do you envision a YAML override for the colors? I would recommend a one-word toggle to pick between the two, with your current definitions as default. Then again, I doubt anyone beyond you and I would care, so it may just be better from a UI perspective to just stick with your defaults.

jdeath · 2026-06-05T11:50:03Z

something like below. then if empty strings, I think will be default color. I made one minor change to script (I messed up the check in API call". should be easy to merge back in or fix after your PR

colorOverride:

red: "XXXXX"
blue: "YYYYY"
yellow: "XXXX"

For the browse script, don't really care or a CLI argument

AESternberg · 2026-06-05T15:06:40Z

Sounds great. If I may suggest, from a user interface point of view, I'd define a couple of keywords for people to use in the YAML (such as 'default', 'bright ', etc. and translate them in the code.Sent from my Galaxy -------- Original message --------From: jdeath ***@***.***> Date: 6/5/26 7:51 AM (GMT-05:00) To: jdeath/CheckRoyalCaribbeanPrice ***@***.***> Cc: AESternberg ***@***.***>, Author ***@***.***> Subject: Re: [jdeath/CheckRoyalCaribbeanPrice] Maintenance refactor - 2nd pass (PR #72) jdeath left a comment (jdeath/CheckRoyalCaribbeanPrice#72) something like below. then if empty strings, I think will be default color. I made one minor change to script (I messed up the check in API call". should be easy to merge back in or fix after your PR colorOverride: red: "XXXXX" blue: "YYYYY" yellow: "XXXX" —Reply to this email directly, view it on GitHub, or unsubscribe.Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS and Android. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

jdeath · 2026-06-05T21:36:23Z

Maybe both. I think fine control is good, with the option of simple providing a theme. So its:
colorOverride: bright or
colorOverride:
-red:YYY

yellow: XXX
blue: zzz

AESternberg · 2026-06-15T02:09:11Z

Hi, jdeath,
It's been several days and this PR is still open. Did you have any more questions or concerns you'd like me to address?

jdeath · 2026-06-15T23:32:05Z

I haven't had a chance to test yet. I will soon

AESternberg · 2026-06-15T23:44:10Z

Understood; it's been a pretty busy time for me as well. As always, feel free to let me know if I can help.

I did refactor Browse as well in the meantime, and made one more change to Check to just rearrange the functions to a more logical grouping (based on what they do). I was going to wait to submit until this one was merged, but can do so if you'd prefer. For my two cents, I think keeping them separate is better just from a "complexity of review" standpoint, but it's ultimately your call.

jdeath · 2026-06-15T23:57:24Z

Still errors and have not had a successful run yet.

Still ntfy on booked cruise no longer for sale. Do not want to notify in that case. Only notify if using watchlist and sold out
code crashes due to bad variable.

ERROR: NameError: name 'watch_item' is not defined
Traceback (most recent call last):
  CheckRoyalCaribbeanPriceAES.py", line 1677, in get_orders
    reservations=getattr(watch_item, 'reservations', [])
                         ^^^^^^^^^^
NameError: name 'watch_item' is not defined

If you can, update:
params = {'brand': brand_code, 'includeCheckin': 'false'}
to
params = {'brand': brand_code, 'includeCheckin': 'true'}

that was a bug that should been fixed before.

I'll look into the browse script later.

AESternberg · 2026-06-16T00:01:49Z

I surprised there are still errors and deeply apologize; I hate wasting your time and really thought I'd caught them all (there must be some execution paths I'm unable to test).

I've not yet pushed the other changes I was talking about, but will do so with that fix.

get_products_graph_all_pages continued on an empty commerceProducts list, contradicting both its own comment ("Breaks out immediately when an API response is empty") and the pagination test's stated expectation. If the API returns an empty success page, the loop fires all remaining pagination requests: a one-product category costs 100 API calls (reproducible via the existing test scenario). Today's API returns None (fragment miss) for exhausted pages, so this is latent rather than active - verified against the live products endpoint, where a real category paginates 12, 10, None - but one schema change away from hammering the API. Treat None and [] both as the terminal page, and assert the call count in the pagination test so the loop behavior is actually verified. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Port of the fix merged into jdeath's main (PR jdeath#74) for issue jdeath#64, where the Browse script was specifically confirmed affected: the RCCL edge servers reject Python's default TLS fingerprint as a bot for some IPs/systems, returning 403 Access Denied even with browser headers. - Import requests from curl_cffi when available; fall back to plain requests when not installed (e.g. iOS), with no behavior change - The request engine passes impersonate="chrome" via IMPERSONATE_ARGS (empty in the fallback, so plain requests sees no unknown kwargs) - The engine success test asserts the kwargs via IMPERSONATE_ARGS so it passes with and without curl_cffi installed (verified in both) curl_cffi is already in requirements.txt on jdeath's main; the Check script port (PR #4 on this fork) adds it here. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

The engine was single-shot: during the minutes-long product sweep, one 504 either exited the whole interactive session (exit_on_fail=True paths) or silently truncated a category's results (pagination treats a failed page as done). Given how often the API throws 504s, that costs real browsing sessions. Mirror the retry design from CheckRoyalCaribbeanPrice's engine: connection-level errors and retryable statuses (429/5xx) retry up to 3 attempts with exponential backoff; definitive HTTP answers like 404 are NOT retried since they signal terminal pagination boundaries and would otherwise add ~6s of pointless backoff at every category end. Existing failure tests updated for retry counts (with time.sleep patched so they stay fast); adds tests for 503-then-recover and for 404 not being retried. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

The 175-point figure came from a Royal Caribbean loyalty PDF that was corrected two days after publication: the single supplement benefit still requires 340 Crown & Anchor points, not Diamond Plus tier alone. jdeath's main already reverted for the same reason (commit 844bd35, "Revert to needing 340 points for reduced single rate"). With the override enabled, members holding 175-339 points get quoted DP340 prices the booking site will not honor, producing Rebook alerts for prices that do not survive checkout. Keeps the diamond_plus_override switch in place (now False) in case RCCL ever makes the 175-point change for real. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

.get(key, 0) only applies the default when the key is absent; a key present with JSON null returns None, and the numeric comparisons downstream (shared points > 0, the dp340 >= 340 eligibility check, and the Club Royale tier computation) then raise TypeError and kill the run. The original script guards these spots explicitly with "is not None" checks and an "Error finding C&A Points" fallback, which suggests null point values do occur in the wild. Add "or 0" to the three point lookups used in comparisons, plus a test with an all-null loyalty payload. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

33 tests pinning down when notifications fire and - just as important - when they must stay silent, across: price direction x booked-vs- watchlist x final-payment window x minimumSavingAlert thresholds (including the exact-boundary case, which uses strict <) x per-night saving totalization x OBC display x refundable/all-included fares x missing price data x reservation filtering. Every test asserts on apobj.notify behavior (count and body), not just log text. Validated by mutation testing: eight deliberate logic breaks (threshold boundary flips, removing the past-final-payment gate, dropping per-night totalization, inverting the reservation filter, deleting the notify call, removing the shipboard price fallback) were each caught by a failing test. Coverage: get_new_order_price 47% -> 92%, get_cruise_price 68% -> 89%. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Guard against None responses when API retries are exhausted

Bail out instead of sending a false Rebook alert when fare data is missing

Minor hardening: login timeout and found_items as a set

Use browser TLS impersonation via curl_cffi to fix 403 Access Denied

Support ${ENV_VAR} expansion for config values

tecmage · 2026-07-15T02:38:42Z

I'm done now, I swear.... Unless you have any other ideas.

AESternberg · 2026-07-15T02:45:51Z

Just what I'd mentioned in my comment to @jdeath above

Add requestTimeout config key for a configurable API timeout

Show partial check-in entries in yellow

Strip design-descriptor words from promo names parsed out of filenames

Break pagination on empty product pages instead of continuing

Browse: use browser TLS impersonation via curl_cffi to fix 403 Access Denied

Revert DP340 threshold to 340 points - the 175 figure was a corrected PDF typo

Guard against null loyalty point values from the profile API

Add alert decision matrix tests for the two money functions

Show check-in opening as local date and time, not raw ISO date

Browse: retry transient API failures with backoff in the request engine

- Patched `_execute_api_request` to safely extract status codes from sparse mock exceptions. - Ensured client-side 4xx errors cleanly trigger immediate failure exit/return paths instead of retrying.

AESternberg · 2026-07-15T03:36:56Z

Merged and tested various changes from @tecmage (thanks, brother!). Questions asked previously still pending. #72 (comment)

tecmage · 2026-07-15T13:56:07Z

Just what I'd mentioned in my comment to @jdeath above

I do like the proposed updated formatting and if there's not a room available it makes sense to have it say "No alternative room inventory returned by the booking engine" or something similar.

AESternberg · 2026-07-15T14:41:30Z

Thanks. Since those are effectively design changes, though, I wouldn't make them without jdeath's approval. Sent from my Galaxy -------- Original message --------From: tecmage ***@***.***> Date: 7/15/26 9:57 AM (GMT-05:00) To: jdeath/CheckRoyalCaribbeanPrice ***@***.***> Cc: AESternberg ***@***.***>, Author ***@***.***> Subject: Re: [jdeath/CheckRoyalCaribbeanPrice] Maintenance refactor - 2nd pass (PR #72) tecmage left a comment (jdeath/CheckRoyalCaribbeanPrice#72) Just what I'd mentioned in my comment to @jdeath above I do like the proposed updated formatting and if there's not a room available it makes sense to have it say "No alternative room inventory returned by the booking engine" or something similar. —Reply to this email directly, view it on GitHub, or unsubscribe.Triage notifications, keep track of coding agent tasks and review pull requests on the go with GitHub Mobile for iOS and Android. Download it today! You are receiving this because you authored the thread.Message ID: ***@***.***>

AESternberg added 13 commits April 28, 2026 20:14

Minor cleanup and add 'Cabin' label to room prints

be90f84

Merge branch 'main' of https://github.com/AESternberg/CheckRoyalCarib…

9f782de

…beanPrice

Move obsolete/unused functions to the bottom

3ab7c37

Add print_response debug function to facilitate API debug

7638679

Merge branch 'main' of https://github.com/AESternberg/CheckRoyalCarib…

67c2753

…beanPrice

Merge branch 'main' of https://github.com/AESternberg/CheckRoyalCarib…

058a1d3

…beanPrice

Dead/Obsolte/Unused function section header

faabea8

Maintenance refactor

25d7109

Refactor debug

e72eae3

Enhanced comments; replaced print with log

f487c6b

Merge branch 'main' of https://github.com/AESternberg/CheckRoyalCarib…

a297bfc

…beanPrice

Review comments

c753c7e

More review comments

a57cc71

Addressing more code review comments and answered questions

b3952fe

AESternberg added 3 commits June 15, 2026 22:54

Architectural/maintenance refactor

4dfad28

Re-arrange functions for logical organization (maintenance); bug fixes

f433b1f

Create new pytest/mock test suites

d98a8a3

tecmage and others added 11 commits July 14, 2026 18:50

Merge pull request #1 from tecmage/aes-retry-none-guards

82de574

Guard against None responses when API retries are exhausted

Merge pull request #2 from tecmage/aes-fare-fallback

09aa991

Bail out instead of sending a false Rebook alert when fare data is missing

Merge pull request #3 from tecmage/aes-minor-hardening

ef45f04

Minor hardening: login timeout and found_items as a set

Merge pull request #4 from tecmage/aes-port-curl-cffi

c66244c

Use browser TLS impersonation via curl_cffi to fix 403 Access Denied

Merge pull request #5 from tecmage/aes-port-env-vars

69c0065

Support ${ENV_VAR} expansion for config values

Merge branch 'main' into aes-port-request-timeout

963d490

AESternberg added 13 commits July 14, 2026 22:46

Merge pull request #6 from tecmage/aes-port-request-timeout

97f7b5d

Add requestTimeout config key for a configurable API timeout

Merge pull request #7 from tecmage/aes-partial-checkin-yellow

215675f

Show partial check-in entries in yellow

Merge pull request #9 from tecmage/aes-promo-name-cleanup

8a2c3fb

Strip design-descriptor words from promo names parsed out of filenames

Merge pull request #10 from tecmage/aes-browse-pagination-fix

d390a2e

Break pagination on empty product pages instead of continuing

Merge pull request #11 from tecmage/aes-browse-curl-cffi

37b89b5

Browse: use browser TLS impersonation via curl_cffi to fix 403 Access Denied

Merge pull request #13 from tecmage/aes-dp340-revert

5b465e3

Revert DP340 threshold to 340 points - the 175 figure was a corrected PDF typo

Merge pull request #14 from tecmage/aes-loyalty-none-guard

2a01501

Guard against null loyalty point values from the profile API

Merge pull request #15 from tecmage/aes-alert-matrix-tests

c03f373

Add alert decision matrix tests for the two money functions

Merge branch 'main' into aes-checkin-local-time

46264fc

Merge pull request #8 from tecmage/aes-checkin-local-time

800c3fa

Show check-in opening as local date and time, not raw ISO date

Merge branch 'main' into aes-browse-retries

4468ac2

Merge pull request #12 from tecmage/aes-browse-retries

3c56f78

Browse: retry transient API failures with backoff in the request engine

Fix test regressions in Browse by refining HTTPError mock handling

2792e7b

- Patched `_execute_api_request` to safely extract status codes from sparse mock exceptions. - Ensured client-side 4xx errors cleanly trigger immediate failure exit/return paths instead of retrying.

Conversation

AESternberg commented May 29, 2026

Description

Key Enhancements & Structural Themes

Pending Questions (Technical Debt)

Uh oh!

jdeath commented May 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

AESternberg commented May 31, 2026

Uh oh!

AESternberg commented May 31, 2026

Uh oh!

jdeath commented May 31, 2026

Uh oh!

AESternberg commented May 31, 2026

Uh oh!

jdeath commented Jun 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

AESternberg commented Jun 5, 2026 via email

Uh oh!

jdeath commented Jun 5, 2026

Uh oh!

AESternberg commented Jun 15, 2026

Uh oh!

jdeath commented Jun 15, 2026

Uh oh!

AESternberg commented Jun 15, 2026

Uh oh!

jdeath commented Jun 15, 2026

Uh oh!

AESternberg commented Jun 16, 2026

Uh oh!

tecmage commented Jul 15, 2026

Uh oh!

AESternberg commented Jul 15, 2026

Uh oh!

AESternberg commented Jul 15, 2026

Uh oh!

tecmage commented Jul 15, 2026

Uh oh!

AESternberg commented Jul 15, 2026 via email

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

jdeath commented May 31, 2026 •

edited

Loading

jdeath commented Jun 5, 2026 •

edited

Loading